{"templateName":"blog-page","allowedRenditionsWidth":["320","480","640","768","960","1200","1440","1920"],"cssClassNames":"blog-page page basicpage summit-page","canonicalLink":"https://www.snowflake.com/en/blog/engineering/snowflake-coco-cowork-token-spend-efficiency/","robotsTags":["index","follow"],"language":"en","description":"Learn how Snowflake CoCo and CoWork cut AI agent token spend by up to 66% while boosting data analysis accuracy through context compaction and Cortex Sense.","title":"Snowflake CoCo & CoWork: Token & Intelligence Efficiency","analyticsPageType":"homepage","analyticsCategory":"general","analyticsSubCategory":"","excludeFromAnalytics":false,":type":"snowflake-site/components/structure/page",":mappedPath":"/en/blog/engineering/snowflake-coco-cowork-token-spend-efficiency/","analyticsEnabled":true,"coveoConfig":{"apiKey":"xx335921a6-2a0a-40f2-a167-e390b4766c3d","pipeline":"snowflake.com","searchHub":"snowflake.com","organizationId":"snowflakecomputingproduction8neljofn"},"isPasswordProtected":false,"analyticsContentTags":["snowflake-site:taxonomy/blog/engineering-blog/data-engineering","snowflake-site:taxonomy/blog/engineering-blog/gen-ai"],"analyticsDebugMode":false,"analyticsData":{"excludeFromAnalytics":false,"subCategory":"","pageType":"homepage","templateName":"blog-page","siteName":"snowflake","pageUrl":"/content/snowflake-site/global/en/blog/engineering/snowflake-coco-cowork-token-spend-efficiency","language":"en","category":"general","pageName":"Intelligence Efficiency in Action: Lower Cost per Trusted Outcome with Snowflake CoCo and CoWork","contentTags":["snowflake-site:taxonomy/blog/engineering-blog/data-engineering","snowflake-site:taxonomy/blog/engineering-blog/gen-ai"]},":items":{"root":{"columnClassNames":{"experiencefragment-banner":"aem-GridColumn aem-GridColumn--default--12","experiencefragment-sub-header":"aem-GridColumn aem-GridColumn--default--12","experiencefragment-pre-footer":"aem-GridColumn aem-GridColumn--default--12","experiencefragment-header":"aem-GridColumn aem-GridColumn--default--12","markup_editor-table":"aem-GridColumn aem-GridColumn--default--12","responsivegrid":"aem-GridColumn aem-GridColumn--default--12","experiencefragment-footer":"aem-GridColumn aem-GridColumn--default--12","markup_editor":"aem-GridColumn aem-GridColumn--default--12","container_47873732":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","columnCount":12,":items":{"experiencefragment-banner":{"id":"experiencefragment-9c6cd14cc9","localizedFragmentVariationPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/pushdown-banner/pushdown-banner-blank/jcr:content","configured":true,":type":"snowflake-site/components/experiencefragment","xfModelPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/pushdown-banner/pushdown-banner-blank.xfmodel.json"},"experiencefragment-header":{"id":"experiencefragment-bbfae48b34","localizedFragmentVariationPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/mega-nav-header/master/jcr:content","configured":true,":type":"snowflake-site/components/experiencefragment","xfModelPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/mega-nav-header/master.xfmodel.json","languageNavPath":"/content/snowflake-site/global/en/blog/engineering/snowflake-coco-cowork-token-spend-efficiency.languagenav.json","appliedCssClassNames":"snowflake-sticky-nav-host"},"experiencefragment-sub-header":{"id":"experiencefragment-0d9d542557","localizedFragmentVariationPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/sub-navigation/engineering-blog-sub-nav/jcr:content","configured":true,":type":"snowflake-site/components/experiencefragment","xfModelPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/sub-navigation/engineering-blog-sub-nav.xfmodel.json"},"responsivegrid":{"columnClassNames":{"container_breadcrumb":"aem-GridColumn aem-GridColumn--default--12","container_main_content":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","columnCount":12,":items":{"container_breadcrumb":{"columnClassNames":{"breadcrumb":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","layout":"RESPONSIVE_GRID","columnCount":12,"id":"blog-page-breadcrumb-indentation",":type":"snowflake-site/components/container",":items":{"breadcrumb":{"id":"breadcrumb-0d01472d20","breadcrumbItems":[{"title":"Blog","path":"/en/blog/engineering/","active":false},{"title":"Gen AI","path":"/en/blog/engineering/gen-ai/","active":false},{"title":"Intelligence Efficiency in Action: Lower Cost per Trusted Outcome with Snowflake CoCo and CoWork","path":"/en/blog/engineering/snowflake-coco-cowork-token-spend-efficiency/","active":false}],":type":"snowflake-site/components/blog/breadcrumb"}},":itemsOrder":["breadcrumb"],"appliedCssClassNames":"snowflake-container"},"container_main_content":{"columnClassNames":{"flexible_column_container":"aem-GridColumn aem-GridColumn--default--12","related_content":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","layout":"RESPONSIVE_GRID","columnCount":12,"id":"main-content",":type":"snowflake-site/components/container",":items":{"flexible_column_container":{"id":"flexible-column-container-b9aec35fe7","propertiesId":"snowflake-blog-template-main-container","type":"2-column-60-40","alignColumns":"top","containerMaxWidth":"extra-large","topPadding":"none","bottomPadding":"none","spaceBetween":"none","reverseOnMobile":true,"carouselOnMobile":false,"backgroundImageOption":"none","flexible_column_content_container_1":{"layout":"SIMPLE","id":"container-43a2ccb4c3",":type":"snowflake-site/components/flexible-column-container/flexible-column-content-container",":items":{"container_hero":{"columnClassNames":{"blog_hero":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","layout":"RESPONSIVE_GRID","columnCount":12,"id":"container-eaf1e86e56",":type":"snowflake-site/components/container",":items":{"blog_hero":{"id":"blog-hero-83813c3d33","showClaude":true,"showChatGpt":true,"authors":[{"authorImage":{"id":"image-1ed29066df","height":"800","isLcpImage":false,"src":"https://www.snowflake.com/adobe/dynamicmedia/deliver/dm-aid--249da901-4810-48b7-ab40-99208c5e3b73/default-author-image.png?preferwebp=true&quality=85","alt":"Snowflake AI Research","lazyEnabled":true,"width":"800",":type":"snowflake-site/components/image"},"authorCta":{"id":"button-dfe3790fc0","showOutboundIcon":false,"buttonLink":{"valid":true,"url":"/en/blog/authors/snowflake-ai-research/"},"linkTargetContentType":"DOCUMENT_LEARN",":type":"snowflake-site/components/button","linkType":"SNOWFLAKE_INTERNAL","text":"Snowflake AI Research"}}],"image":{"id":"image-6646839bee","height":"720","isLcpImage":false,"src":"https://www.snowflake.com/adobe/dynamicmedia/deliver/dm-aid--85d72614-270b-4273-9e68-6bb4ddc3a612/sf-eng-blog-ml-3.png?preferwebp=true&quality=85","lazyEnabled":true,"width":"1680",":type":"snowflake-site/components/image"},"timeToRead":"11","publicationDate":"AUG 21, 2026","tag":{"tagText":"Gen AI","tagColor":"#29B5E8"},"title":{"lines":["Intelligence Efficiency in Action: Lower Cost per Trusted Outcome with Snowflake CoCo and CoWork"],"type":"heading2",":type":"snowflake-site/components/title-v2"},":type":"snowflake-site/components/blog/blog-hero"}},":itemsOrder":["blog_hero"]},"responsivegrid_content":{"columnClassNames":{"image":"aem-GridColumn aem-GridColumn--default--12","blog_text_29009044":"aem-GridColumn aem-GridColumn--default--12","blog_text_1416535635":"aem-GridColumn aem-GridColumn--default--12","blog_text_1802675966":"aem-GridColumn aem-GridColumn--default--12","blog_text_665354999":"aem-GridColumn aem-GridColumn--default--12","blog_text_214029945":"aem-GridColumn aem-GridColumn--default--12","blog_text_1267016713":"aem-GridColumn aem-GridColumn--default--12","blog_text_1921458890":"aem-GridColumn aem-GridColumn--default--12","blog_text_799097899":"aem-GridColumn aem-GridColumn--default--12","blog_text_1162420559":"aem-GridColumn aem-GridColumn--default--12","image_1722941509":"aem-GridColumn aem-GridColumn--default--12","image_806067150":"aem-GridColumn aem-GridColumn--default--12","blog_text_1203348665":"aem-GridColumn aem-GridColumn--default--12","blog_text":"aem-GridColumn aem-GridColumn--default--12","blog_text_2118694130":"aem-GridColumn aem-GridColumn--default--12","blog_text_102837410":"aem-GridColumn aem-GridColumn--default--12","blog_text_1683766248":"aem-GridColumn aem-GridColumn--default--12","blog_text_537712058":"aem-GridColumn aem-GridColumn--default--12","image_1179192190":"aem-GridColumn aem-GridColumn--default--12","blog_text_1671380790":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","columnCount":12,"appliedCssClassNames":"snowflake-layout-container-inner-padding-small",":items":{"blog_text":{"id":"blog-text-15411e93f3","text":"\u003Cp\u003E\u003Ca href=\"https://www.snowflake.com/en/blog/ai-intelligence-efficiency-dynamic-model-routing/\" target=\"_blank\" rel=\"noopener noreferrer\"\u003EIntelligence efficiency\u003C/a\u003E measures how effectively a company turns compute, models, data and context into business impact. The right metric is cost per outcome you can trust, not cost per attempt.\u003C/p\u003E\r\n\u003Cp\u003EProduction-grade data analysis tasks require agents to reason about business logic, handle edge cases, perform exploratory data analysis and navigate large warehouses. These sessions tend to be long, with large tool outputs that accumulate in context over many turns. As agentic workflows move into production, intelligence efficiency becomes a significant concern. A key element towards achieving intelligence efficiency in data analysis tasks is to reduce cost while maintaining quality.\u003C/p\u003E\r\n\u003Cp\u003EAgents such as \u003Ca href=\"https://docs.snowflake.com/en/user-guide/cortex-code/cortex-code\" target=\"_blank\" rel=\"noopener noreferrer\"\u003ESnowflake CoCo\u003C/a\u003E and \u003Ca href=\"https://docs.snowflake.com/en/user-guide/snowflake-cortex/snowflake-cowork\" target=\"_blank\" rel=\"noopener noreferrer\"\u003ESnowflake CoWork\u003C/a\u003E can tackle deep data analysis sessions involving hundreds of model calls. In a baseline agent implementation, each call would re-read the full conversation history, reason over it and then produce output, so every token added to that history is paid for again on each turn that follows. This form of \u003Ci\u003Econtext bloat\u003C/i\u003E increases cost \u003Ci\u003Eand\u003C/i\u003E also makes it harder for the model to attend to what actually matters, a failure mode commonly called \u003Ci\u003Econtext rot\u003C/i\u003E. Both pressures grow with session length, which is increasingly common in data analysis workloads. So, we set ourselves a challenge:\u003C/p\u003E\r\n\u003Cp\u003E\u003Cb\u003E\u003Ci\u003ECould we maintain or improve CoCo and CoWork's data analysis quality while reducing token spend?\u003C/i\u003E\u003C/b\u003E\u003C/p\u003E\r\n\u003Cp\u003EOur key findings are two-fold:\u003C/p\u003E\r\n\u003Cul\u003E\r\n\u003Cli\u003E\u003Cb\u003ELoading skills on demand, bundling tools dispatch and compacting tool output significantly cuts cost and raises quality.\u003C/b\u003E Tool schemas, skill catalogs, previous tool outputs and growing history all re-enter the prompt whether or not the current prompt/step needs them. CoCo's harness dramatically reduces the context by \u003Ci\u003Eloading tools and skills on demand\u003C/i\u003E, \u003Ci\u003Ebundling independent tool calls into a single dispatch and compacting tool output before it enters context.\u003C/i\u003E The savings compound: Because the harness preserves quality while cutting tokens, a smaller model on CoCo can match or beat a larger model on a conventional harness.\u003Cul\u003E\r\n\u003Cli\u003EOn an internal agentic SQL-fixing benchmark inspired by production workloads: CoCo on Sonnet 5 outperformed Claude Code on Opus 5 — a smaller model on the stronger harness — with success on three independent trials (Pass\u003Csup\u003E3\u003C/sup\u003E) of 86% vs 72% while cost per trial fell 33% at the same time: $0.298 for CoCo on Sonnet 5 vs $0.446 for Claude Code on Opus 5.\u003C/li\u003E\r\n\u003Cli\u003EThe pattern holds for a broader benchmark of general Snowflake workloads. CoCo on Sonnet 5 matched the reliability of Claude Code on Opus 5 — Pass\u003Csup\u003E3\u003C/sup\u003E of 60% vs 58% while cost per trial fell 45% at the same time: $0.451 for CoCo on Sonnet 5 vs $0.821 for Claude Code on Opus 5.\u003C/li\u003E\r\n\u003C/ul\u003E\r\n\u003C/li\u003E\r\n\u003Cli\u003E\u003Cb\u003EProviding semantic context upfront dramatically increases accuracy while cutting cost.\u003C/b\u003E The idea here is very simple: If we provide the agent necessary context that it would otherwise have to discover, it will not spend turns and tokens on discovering them at query time (for example, in scanning schemas or inferring relationships). Therefore by grounding agents in the right context provided by \u003Ca href=\"https://www.snowflake.com/en/blog/enterprise-ai-agents-grounded-context/\" target=\"_blank\" rel=\"noopener noreferrer\"\u003ECortex Sense (private preview)\u003C/a\u003E we expect to improve both token efficiency and accuracy. To this end, we created a set of 58 internal questions spanning Snowflake's own product analytics domain. To answer these questions correctly, the agent needs to know the right tables from within Snowflake's massive internal data estate as well as the right analytics and business context (for example, to filter out some rows and to perform the right form of aggregations).\u003Cul\u003E\r\n\u003Cli\u003EWe ran a simple Pass@1 evaluation comparing two systems: a) CoWork using Cortex Sense and b) Claude Code connecting to Snowflake via Snowflake MCP. We found that the former achieved \u003Cb\u003E86.3% accuracy at $0.59 per query\u003C/b\u003E on this benchmark, compared to \u003Cb\u003E24.1% accuracy at $1.76 per query\u003C/b\u003E for the latter — \u003Cb\u003Ea 66% cost reduction with a 3.6× quality gain\u003C/b\u003E.\u003C/li\u003E\r\n\u003C/ul\u003E\r\n\u003C/li\u003E\r\n\u003C/ul\u003E\r\n\u003Cp\u003ELet's examine these two findings in more detail.\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_1162420559":{"id":"blog-text-dd5bfe6662","text":"\u003Ch2\u003ELoading skills on demand, bundling tools dispatch and compacting tool output significantly cuts cost and raises quality\u003C/h2\u003E\r\n\u003Cp\u003EConsider the following query: \u003Ci\u003EOver the last 30 days, identify the most common CLI failures. Quantify their frequency and cost from warehouse telemetry, correlate relevant Slack and Jira discussions and locate the responsible code paths. Deduplicate retries and cite the evidence.\u003C/i\u003E\u003C/p\u003E\r\n\u003Cp\u003ETo answer this query, harnesses supplement the user prompt with tool and skill schemas: descriptions of what each tool does, sample usage patterns and the input/output shape it expects. The agent plans against this catalog, calling into warehouse telemetry, Slack and Jira, processing what comes back, and synthesizing over the accumulated context to produce a final answer.\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"image":{"id":"image-acff7bdb1b","height":"853","isLcpImage":true,"src":"https://www.snowflake.com/adobe/dynamicmedia/deliver/dm-aid--fe27925f-8237-4848-aa71-15a2ede90d61/figure-1.-intelligence-efficiency-in-action--reducing-the-token-spend-of-snowflake-coco-and-cowork.png?preferwebp=true&quality=85","alt":"Figure 1.","lazyEnabled":true,"width":"2048",":type":"snowflake-site/components/image"},"blog_text_1921458890":{"id":"blog-text-01aacbd997","text":"\u003Cp\u003E\u003Csmall\u003E\u003Ci\u003EFigure 1. A conventional agentic loop on a multi-source investigation. The full tool and skill catalog loads upfront whether or not the task needs it, and each raw result goes into model context, where it accumulates and is paid for again on every turn that follows.\u003C/i\u003E\u003C/small\u003E\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_799097899":{"id":"blog-text-44612ebeae","text":"\u003Cp\u003E \u003C/p\u003E\r\n\u003Cp\u003EWhen we looked closer at internal CoCo trajectories, we identified several areas for improvement.\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"image_1179192190":{"id":"image-eed4993707","height":"512","isLcpImage":false,"src":"https://www.snowflake.com/adobe/dynamicmedia/deliver/dm-aid--d95ae066-15db-49a8-85cd-2899cba6b37b/figure-2.-intelligence-efficiency-in-action--reducing-the-token-spend-of-snowflake-coco-and-cowork.png?preferwebp=true&quality=85","alt":"Figure 2.","lazyEnabled":true,"width":"2048",":type":"snowflake-site/components/image"},"blog_text_102837410":{"id":"blog-text-ee44dcf42d","text":"\u003Cp\u003E\u003Csmall\u003E\u003Ci\u003EFigure 2. Each lever in isolation: on-demand loading removes ~79k tokens of idle schema and skill context (exact savings vary by setup); Programmatic Tool Calling collapses three independent retrievals into a single round trip; offload and lossless compression hand the model a reference, not the data.\u003C/i\u003E\u003C/small\u003E\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_2118694130":{"id":"blog-text-da8a969a0a","text":"\u003Cp\u003E \u003C/p\u003E\r\n\u003Cp\u003E\u003Cb\u003ELoad on demand:\u003C/b\u003E Only a fraction of the registered catalog is relevant to any one task. A query like this one needs a Slack search tool, a Jira search tool and a Snowflake query tool — not the other dozens of tools a general-purpose harness carries for tasks it isn't running. Loading every schema upfront means the agent pays context for tools it will never call, and the cost scales with the size of the catalog, not the size of the task. CoCo instead exposes a tool search tool: the agent describes what it needs, and the matching tool definitions are retrieved on demand instead of registered wholesale at the start of the session. Skill loading works the same way — CoCo delegates it to a dedicated sidecar that identifies the skills relevant to the current step and loads only those. Together, these mechanisms shrink the idle context window by ~79k tokens for Snowflake internal users (exact numbers vary depending on the CoCo setup for the individual) while leaving the full catalog of tools and skills available to draw on.\u003C/p\u003E\r\n\u003Cp\u003E\u003Cb\u003EBundled dispatch:\u003C/b\u003E Once the agent knows which tools to call, calling them one at a time wastes turns. A harness that queries Slack, then waits for the model to process the result, then queries Jira, then waits again, then queries the warehouse, pays for a full context re-read at each step even though none of the three calls depends on the others' output. Programmatic Tool Calling lets CoCo bundle independent calls, like the Slack search, Jira search, and warehouse query this task needs, into a single turn: The harness dispatches all three, collects the results and returns to the model once instead of three times.\u003C/p\u003E\r\n\u003Cp\u003E\u003Cb\u003ETool output compaction:\u003C/b\u003E Tool output is also processed before it enters context, rather than passed through raw. Oversized results, like a full warehouse telemetry pull or a long Jira thread, are offloaded and held as variables in a persistent Python REPL: the agent gets a reference to the data instead of the data itself, and can retrieve, filter or aggregate it later without re-reading the full payload on every subsequent turn. SQL results are losslessly compressed by exploiting the shape of the data itself (repeated column headers, predictable types, redundant formatting), preserving every value the model needs to reason over while cutting what it doesn't. What the model sees at each turn is a compact payload with pointers into the REPL, sampled and reasoned over as needed to produce the final answer, deduplicated retries and all.\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"image_1722941509":{"id":"image-f404881325","height":"973","isLcpImage":false,"src":"https://www.snowflake.com/adobe/dynamicmedia/deliver/dm-aid--5756cba5-9efa-4da6-a734-1050b53c9d06/figure-3.-intelligence-efficiency-in-action--reducing-the-token-spend-of-snowflake-coco-and-cowork.png?preferwebp=true&quality=85","alt":"Figure 3.","lazyEnabled":true,"width":"2048",":type":"snowflake-site/components/image"},"blog_text_29009044":{"id":"blog-text-538e52fb2d","text":"\u003Cp\u003E\u003Csmall\u003E\u003Ci\u003EFigure 3. The same investigation under CoCo. Discovery, retrieval and reduction run below the context boundary in a persistent Python REPL; one compact payload with pointers crosses into model context, and the window the baseline carried (dashed) is never built.\u003C/i\u003E\u003C/small\u003E\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_1203348665":{"id":"blog-text-053290d371","text":"\u003Cp\u003E \u003C/p\u003E\r\n\u003Cp\u003E\u003Cb\u003ESQL-fixing benchmark:\u003C/b\u003E Combined with CoCo's inherent data analysis capabilities, the token savings let a smaller model carry the task: on our internal SQL-fixing benchmark, CoCo on Sonnet 5 beats Claude Code on Opus 5 on pass consistency (86% vs 72% Pass\u003Csup\u003E3\u003C/sup\u003E) at 33% lower cost per trial ($0.298 vs $0.446). At the matched model, CoCo leads by 26 points on Opus 5 and 42 on Sonnet 5.\u003C/p\u003E\r\n\u003Cp\u003E \u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_665354999":{"id":"blog-text-00e451fda7","text":"\u003Cp\u003E\u003Cb\u003ETroubleshooting Snowflake SQL Dataset:\u003C/b\u003E\u003C/p\u003E\r\n\u003Ctable\u003E\r\n\u003Cthead\u003E\u003Ctr\u003E\u003Cth\u003EHarness\u003C/th\u003E\r\n\u003Cth\u003EModel\u003C/th\u003E\r\n\u003Cth\u003EPass\u003Csup\u003E3\u003C/sup\u003E\u003C/th\u003E\r\n\u003Cth\u003ECost per trial ($)\u003C/th\u003E\r\n\u003C/tr\u003E\u003C/thead\u003E\u003Ctbody\u003E\u003Ctr\u003E\u003Ctd\u003ESnowflake CoCo (optimized)\u003C/td\u003E\r\n\u003Ctd\u003EOpus 5\u003C/td\u003E\r\n\u003Ctd\u003E98.0%\u003C/td\u003E\r\n\u003Ctd\u003E0.404\u003C/td\u003E\r\n\u003C/tr\u003E\u003Ctr\u003E\u003Ctd\u003E \u003C/td\u003E\r\n\u003Ctd\u003ESonnet 5\u003C/td\u003E\r\n\u003Ctd\u003E86.0%\u003C/td\u003E\r\n\u003Ctd\u003E0.298\u003C/td\u003E\r\n\u003C/tr\u003E\u003Ctr\u003E\u003Ctd\u003EClaude Code\u003C/td\u003E\r\n\u003Ctd\u003EOpus 5\u003C/td\u003E\r\n\u003Ctd\u003E72.0%\u003C/td\u003E\r\n\u003Ctd\u003E0.446\u003C/td\u003E\r\n\u003C/tr\u003E\u003Ctr\u003E\u003Ctd\u003E \u003C/td\u003E\r\n\u003Ctd\u003ESonnet 5\u003C/td\u003E\r\n\u003Ctd\u003E44.0%\u003C/td\u003E\r\n\u003Ctd\u003E0.146\u003C/td\u003E\r\n\u003C/tr\u003E\u003C/tbody\u003E\u003C/table\u003E\r\n\u003Cp\u003E\u003Ci\u003EResults from the internal SQL-fixing benchmark (Pass\u003Csup\u003E3\u003C/sup\u003E over three repeated trials). CoCo leads at every matched model; on Sonnet 5 it exceeds Claude Code on Opus 5 at 33% lower cost. In these runs, Claude Code on Sonnet 5 returned fixes without executing them against the warehouse — low cost per trial, but only 44% held up across repeats.\u003C/i\u003E\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_1671380790":{"id":"blog-text-c947bd404c","text":"\u003Cp\u003E \u003C/p\u003E\r\n\u003Cp\u003EThe Sonnet 5 pair is the clearest illustration of why cost per trial alone misleads. Claude Code's $0.146 attempts to fix the SQL in a single shot without checking that the fix works by running the SQL and then iterating if needed. CoCo's additional 15 cents buys execution against the warehouse and a 42-point Pass\u003Csup\u003E3\u003C/sup\u003E reliability gap. This is the sense in which intelligence efficiency is cost per outcome you can trust, not cost per attempt.\u003C/p\u003E\r\n\u003Cp\u003E\u003Cb\u003EDiverse Snowflake Analytics benchmark:\u003C/b\u003E A separate benchmark of diverse Snowflake workloads, spanning large-scale data migrations, Streamlit analytics dashboards and more, provides similar findings. At matched model and effort, CoCo improves Pass\u003Csup\u003E3\u003C/sup\u003E by 10 points on Opus 5 (68% vs 58%) and 12 on Sonnet 5 (60% vs 48%) — and CoCo on Sonnet 5 delivers the reliability of Claude Code on Opus 5 (60% vs 58%) at 45% lower cost per trial ($0.451 vs $0.821).\u003C/p\u003E\r\n\u003Cp\u003E \u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_214029945":{"id":"blog-text-839aa0dff6","text":"\u003Cp\u003E\u003Cb\u003EGeneral Snowflake Workloads Dataset:\u003C/b\u003E\u003C/p\u003E\r\n\u003Ctable\u003E\r\n\u003Cthead\u003E\u003Ctr\u003E\u003Cth\u003EHarness\u003C/th\u003E\r\n\u003Cth\u003EModel\u003C/th\u003E\r\n\u003Cth\u003EPass\u003Csup\u003E3\u003C/sup\u003E\u003C/th\u003E\r\n\u003Cth\u003ECost per trial ($)\u003C/th\u003E\r\n\u003C/tr\u003E\u003C/thead\u003E\u003Ctbody\u003E\u003Ctr\u003E\u003Ctd\u003ESnowflake CoCo (optimized)\u003C/td\u003E\r\n\u003Ctd\u003EOpus 5\u003C/td\u003E\r\n\u003Ctd\u003E68.0%\u003C/td\u003E\r\n\u003Ctd\u003E1.262\u003C/td\u003E\r\n\u003C/tr\u003E\u003Ctr\u003E\u003Ctd\u003E \u003C/td\u003E\r\n\u003Ctd\u003ESonnet 5\u003C/td\u003E\r\n\u003Ctd\u003E63.0%\u003C/td\u003E\r\n\u003Ctd\u003E0.394\u003C/td\u003E\r\n\u003C/tr\u003E\u003Ctr\u003E\u003Ctd\u003EClaude Code\u003C/td\u003E\r\n\u003Ctd\u003EOpus 5\u003C/td\u003E\r\n\u003Ctd\u003E58.0%\u003C/td\u003E\r\n\u003Ctd\u003E0.821\u003C/td\u003E\r\n\u003C/tr\u003E\u003Ctr\u003E\u003Ctd\u003E \u003C/td\u003E\r\n\u003Ctd\u003ESonnet 5\u003C/td\u003E\r\n\u003Ctd\u003E48.0%\u003C/td\u003E\r\n\u003Ctd\u003E0.287\u003C/td\u003E\r\n\u003C/tr\u003E\u003C/tbody\u003E\u003C/table\u003E\r\n\u003Cp\u003E\u003Ci\u003ECost–quality frontier on general Snowflake workloads (Pass\u003Csup\u003E3\u003C/sup\u003E over three trials). At matched model and effort, CoCo improves reliability by 10 points on Opus 5 and 12 on Sonnet 5 over Claude Code, and \u003Cb\u003ECoCo on Sonnet 5 matches Claude Code on Opus 5 at 45% lower cost per trial\u003C/b\u003E. CoCo will be rolling out an auto model selector in order for adjustable model choice per workload in order to achieve these cost efficiency wins without user explicit control.\u003C/i\u003E\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_1683766248":{"id":"blog-text-7a4011f1c4","text":"\u003Cp\u003E \u003C/p\u003E\r\n\u003Cp\u003EBecause the harness shifts the whole cost–quality frontier, model and effort become a per-task choice rather than a global one: a medium-effort Sonnet 5 configuration already reaches 57% Pass\u003Csup\u003E3\u003C/sup\u003E at $0.284 per trial.\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_1416535635":{"id":"blog-text-a6adc2e4db","text":"\u003Ch2\u003EProviding semantic context upfront dramatically increases accuracy while cutting cost.\u003C/h2\u003E\r\n\u003Cp\u003EWhen an agent operates on undocumented data, it resorts to expensive exploration: inspecting tables, reading column metadata and inferring relationships between entities. In our experience, for most enterprises, manually curated semantic views cover roughly 5% of tables. The remaining 95% is where agents spend tokens discovering context.\u003C/p\u003E\r\n\u003Cp\u003E\u003Ca href=\"https://www.snowflake.com/en/blog/enterprise-ai-agents-grounded-context/\" target=\"_blank\" rel=\"noopener noreferrer\"\u003ECortex Sense (private preview)\u003C/a\u003E is designed to reduce this by automatically building semantic understanding from signals the business already produces: past analyst queries, dbt transformations, BI metrics and governed definitions. It ranks conflicting signals by relevance, authority and freshness, surfaces unresolved conflicts to human reviewers, and continuously refreshes as the data estate evolves. This produces runtime context spanning the data estate with little to no manual curation required, working alongside semantic views, which remain authoritative for governed contexts, to cover the undocumented long tail.\u003C/p\u003E\r\n\u003Cp\u003EOn internal benchmarks, CoWork with Cortex Sense improves accuracy from 24.1% to 86.3% and reduces cost from $1.76 to $0.59 per query (a 66% reduction) compared to an ungrounded frontier agent.\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"image_806067150":{"id":"image-282b178e79","height":"576","isLcpImage":false,"src":"https://www.snowflake.com/adobe/dynamicmedia/deliver/dm-aid--3f8b8616-f199-425c-956a-e46aee0d82bb/figure-4.-intelligence-efficiency-in-action--reducing-the-token-spend-of-snowflake-coco-and-cowork.png?preferwebp=true&quality=85","alt":"Figure 4.","lazyEnabled":true,"width":"1920",":type":"snowflake-site/components/image"},"blog_text_537712058":{"id":"blog-text-4b7d8964a5","text":"\u003Cp\u003EAgents receive scoped, grounded context before the session begins, reducing the need to inspect tables and infer relationships at runtime. The previous section makes each task cheaper twice — fewer tokens spent on the work, and a smaller model able to do it. Cortex Sense reduces how much of that work needs to happen in the first place.\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_1267016713":{"id":"blog-text-f74d7770b9","text":"\u003Ch2\u003EGet started\u003C/h2\u003E\r\n\u003Cp\u003EThese optimizations are rolling out across CoCo and CoWork. The token-reduction and grounding levers described above work by default — the harness is designed to keep per-user spend lower without requiring manual intervention. For organizations that need explicit governance, administrators can set \u003Ca href=\"https://docs.snowflake.com/en/user-guide/cortex-code/credit-usage-limit\" target=\"_blank\" rel=\"noopener noreferrer\"\u003Eper-user daily credit limits\u003C/a\u003E across CoCo surfaces (CLI, Desktop, Snowsight) at the account or user level, and track consumption patterns through \u003Ca href=\"https://docs.snowflake.com/en/sql-reference/account-usage/cortex_code_cli_usage_history\" target=\"_blank\" rel=\"noopener noreferrer\"\u003Eusage history views\u003C/a\u003E with per-request granularity by model, user and region.\u003C/p\u003E\r\n\u003Cp\u003ETry \u003Ca href=\"https://www.snowflake.com/en/product/snowflake-coco/\" target=\"_blank\" rel=\"noopener noreferrer\"\u003ECoCo\u003C/a\u003E for data engineering or \u003Ca href=\"https://www.snowflake.com/en/product/snowflake-cowork/\" target=\"_blank\" rel=\"noopener noreferrer\"\u003ECoWork\u003C/a\u003E for analytics.\u003C/p\u003E\r\n\u003Cp\u003E\u003Ci\u003EForward-looking statements: This article contains forward-looking statements, including about our future product offerings, and are not commitments to deliver any product offerings. Actual results and offerings may differ and are subject to known and unknown risk and uncertainties. See our latest 10-Q for more information.\u003C/i\u003E\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"},"blog_text_1802675966":{"id":"blog-text-b59a37fd8a","text":"\u003Chr\u003E\r\n\r\n\u003Cp\u003E\u003Csmall\u003E\u003Csup\u003E1\u003C/sup\u003E Much of this is on by default in CoCo today; sidecar-based skill loading and lossless result compression are coming soon to the product.\u003C/small\u003E\u003C/p\u003E\r\n\u003Cp\u003E\u003Csmall\u003E\u003Csup\u003E2\u003C/sup\u003E Cost optimization features are in either private preview or public preview, depending on the applicable model. \u003Ca href=\"https://www.snowflake.com/en/blog/enterprise-ai-agents-grounded-context/\" target=\"_blank\" rel=\"noopener noreferrer\"\u003ECortex Sense\u003C/a\u003E is currently in private preview.\u003C/small\u003E\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/blog/blog-text"}},":itemsOrder":["blog_text","blog_text_1162420559","image","blog_text_1921458890","blog_text_799097899","image_1179192190","blog_text_102837410","blog_text_2118694130","image_1722941509","blog_text_29009044","blog_text_1203348665","blog_text_665354999","blog_text_1671380790","blog_text_214029945","blog_text_1683766248","blog_text_1416535635","image_806067150","blog_text_537712058","blog_text_1267016713","blog_text_1802675966"],":type":"wcm/foundation/components/responsivegrid"},"responsivegrid_premium_content_banner":{"columnClassNames":{},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","columnCount":12,"appliedCssClassNames":"snowflake-responsive-component-top-padding-medium",":items":{},":itemsOrder":[],":type":"wcm/foundation/components/responsivegrid"},"container_author_chip":{"columnClassNames":{"author_chip":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","layout":"RESPONSIVE_GRID","columnCount":12,"id":"container-18f0cd1657",":type":"snowflake-site/components/container",":items":{"author_chip":{"id":"author-chip-2224998746","title":{"id":"title","type":"heading2","lines":["Learn more about the authors"],":type":"snowflake-site/components/title-v2"},"authors":[{"authorImage":{"id":"image-1ed29066df","height":"800","isLcpImage":false,"src":"https://www.snowflake.com/adobe/dynamicmedia/deliver/dm-aid--249da901-4810-48b7-ab40-99208c5e3b73/default-author-image.png?preferwebp=true&quality=85","alt":"Snowflake AI Research","lazyEnabled":true,"width":"800",":type":"snowflake-site/components/image"},"authorCta":{"id":"button-dfe3790fc0","showOutboundIcon":false,"buttonLink":{"valid":true,"url":"/en/blog/authors/snowflake-ai-research/"},"linkTargetContentType":"DOCUMENT_LEARN",":type":"snowflake-site/components/button","linkType":"SNOWFLAKE_INTERNAL","text":"Snowflake AI Research"}}],":type":"snowflake-site/components/blog/author-chip"}},":itemsOrder":["author_chip"],"appliedCssClassNames":"snowflake-responsive-component-top-padding-medium"},"container_share_article":{"columnClassNames":{"share_article":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","layout":"RESPONSIVE_GRID","columnCount":12,"id":"container-e9d652e078",":type":"snowflake-site/components/container",":items":{"share_article":{"id":"share-article-6696006ae5","linkedInShareUrl":"https://www.linkedin.com/shareArticle?mini=true&url=https%3A%2F%2Fwww.snowflake.com%2Fcontent%2Fsnowflake-site%2Fglobal%2Fen%2Fblog%2Fengineering%2Fsnowflake-coco-cowork-token-spend-efficiency&title=Intelligence+Efficiency+in+Action%3A+Lower+Cost+per+Trusted+Outcome+with+Snowflake+CoCo+and+CoWork","twitterShareUrl":"https://x.com/intent/post?url=https%3A%2F%2Fwww.snowflake.com%2Fcontent%2Fsnowflake-site%2Fglobal%2Fen%2Fblog%2Fengineering%2Fsnowflake-coco-cowork-token-spend-efficiency&text=Intelligence+Efficiency+in+Action%3A+Lower+Cost+per+Trusted+Outcome+with+Snowflake+CoCo+and+CoWork","facebookShareUrl":"https://www.facebook.com/sharer/sharer.php?u=https%3A%2F%2Fwww.snowflake.com%2Fcontent%2Fsnowflake-site%2Fglobal%2Fen%2Fblog%2Fengineering%2Fsnowflake-coco-cowork-token-spend-efficiency",":type":"snowflake-site/components/blog/share-article"}},":itemsOrder":["share_article"],"appliedCssClassNames":"snowflake-responsive-component-top-padding-small"}},":itemsOrder":["container_hero","responsivegrid_content","responsivegrid_premium_content_banner","container_author_chip","container_share_article"]},"flexible_column_content_container_2":{"layout":"SIMPLE","id":"container-7b9f11f4dd",":type":"snowflake-site/components/flexible-column-container/flexible-column-content-container",":items":{"blog_table_of_content":{"id":"blog-table-of-content-0848634809",":type":"snowflake-site/components/blog/blog-table-of-content","tableOfContents":[]}},":itemsOrder":["blog_table_of_content"]},":type":"snowflake-site/components/flexible-column-container","isBlogPage":true,"isActiveTOC":false},"related_content":{"id":"related-content-76bddb7b69","relatedContent":[],":type":"snowflake-site/components/blog/related-content","isBlogPage":true}},":itemsOrder":["flexible_column_container","related_content"],"appliedCssClassNames":"snowflake-container"}},":itemsOrder":["container_breadcrumb","container_main_content"],":type":"wcm/foundation/components/responsivegrid"},"container_47873732":{"additionalClasses":"section--blog-newsletter","columnClassNames":{"flexible_column_cont":"aem-GridColumn aem-GridColumn--default--12"},"gridClassNames":"aem-Grid aem-Grid--12 aem-Grid--default--12","layout":"RESPONSIVE_GRID","columnCount":12,"id":"container-b9c77b96e6",":type":"snowflake-site/components/container",":items":{"flexible_column_cont":{"id":"flexible-column-container-288292b78a","type":"1-column","alignColumns":"top","containerMaxWidth":"extra-large","topPadding":"small","bottomPadding":"none","spaceBetween":"small","reverseOnMobile":false,"carouselOnMobile":false,"propertiesCSSClasses":"section--blog-newsletter","backgroundImageOption":"none","flexible_column_content_container_1":{"layout":"SIMPLE","id":"container-9f37b6633c",":type":"snowflake-site/components/flexible-column-container/flexible-column-content-container",":items":{"marketo_v2":{"id":"marketo-v2-03c31c4d42","marketoForm":{"hidden":null,"formId":"3320","edit":false,"successUrl":null,"script":null,"values":null},"title":{"id":"title","type":"heading3","lines":["Subscribe to our blog newsletter","Get the best, coolest and latest delivered to your inbox each week"],":type":"snowflake-site/components/title-v2"},"partnerCampaignPage":false,"marketoConfigured":true,"formConfigured":true,"munchkinId":"252-RFO-227","serverInstance":"252-RFO-227.mktoweb.com",":type":"snowflake-site/components/form/marketo-v2"},"text":{"id":"text-e6949053bb","additionalClasses":"newsletter-disclaimer","text":"\u003Cp\u003EBy submitting this form, I understand Snowflake will process my personal information in accordance with their Privacy Notice.\u003C/p\u003E\r\n","richText":true,":type":"snowflake-site/components/text"}},":itemsOrder":["marketo_v2","text"]},":type":"snowflake-site/components/flexible-column-container","isBlogPage":true,"isActiveTOC":false}},":itemsOrder":["flexible_column_cont"],"appliedCssClassNames":"snowflake-container"},"experiencefragment-pre-footer":{"id":"experiencefragment-73d233d21d","localizedFragmentVariationPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/get-started-pre-footer/get-started-pre-footer/jcr:content","configured":true,":type":"snowflake-site/components/experiencefragment","xfModelPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/get-started-pre-footer/get-started-pre-footer.xfmodel.json"},"markup_editor":{"id":"markup-editor-a3fe41ff69","title":"Page CSS","cssContent":"@media screen and (min-width:768px){.snowflake-blog-author-chip-wrapper{justify-content:flex-start}.snowflake-blog-related-content-on-blog-page{max-width:1408px;margin-left:auto;margin-right:auto}.snowflake-text{font-family:Lato,sans-serif;font-weight:400;font-size:16px;line-height:24px}}.section--blog-newsletter{max-width:none;width:100%;padding-left:0;padding-right:0;margin-left:0;margin-right:0;margin-bottom:0}.section--blog-newsletter .mktoField{background-color:transparent !important}.section--blog-newsletter\u003E.container{padding-left:0;padding-right:0}@media screen and (min-width:768px){.section--blog-newsletter\u003E.container{padding-left:0;padding-right:0}}.newsletter-disclaimer p{font-size:14px !important}.section--blog-newsletter .snowflake-marketo-form-container{margin-bottom:24px;background-color:#f6f9fa;gap:48px;box-shadow:none}.section--blog-newsletter .snowflake-title p.snowflake-title-line:first-child{font-family:Texta;font-size:24px;line-height:26px;font-weight:700;margin-bottom:4px}.section--blog-newsletter .snowflake-title p.snowflake-title-line{text-transform:none;font-family:\"Lato\",sans-serif;font-size:16px;line-height:24px;font-weight:normal}@media screen and (min-width:1024px){.section--blog-newsletter .snowflake-marketo-form-container{display:flex;justify-content:center}.section--blog-newsletter .snowflake-title .snowflake-title-line{text-align:left}.section--blog-newsletter .snowflake-marketo-form .mktoFormRow:has(\u003E input[type=\"hidden\"]){flex-grow:0}.section--blog-newsletter .snowflake-marketo-form{display:flex;width:50% !important}.section--blog-newsletter .snowflake-marketo-form .mktoButtonRow{flex-grow:0;width:auto !important;margin-left:0;margin-right:0}.section--blog-newsletter .snowflake-marketo-form .mktoFormRow{flex-grow:1}.section--blog-newsletter\u003E.container{padding-left:0;padding-right:0}.section--blog-newsletter .snowflake-marketo-form-title{width:50%;margin-bottom:0 !important}.section--blog-newsletter .center .snowflake-title{align-items:flex-start}}.snowflake-sub-navigation a.snowflake-sub-navigation-primary-link{width:auto !important}.snowflake-blog-hero{align-items:stretch !important}",":type":"snowflake-site/components/markup-editor","isGSAPEnabled":false},"markup_editor-table":{"id":"markup-editor-fd7fb0acc0","title":"Table Styling CSS","cssContent":"#snowflake-blog-template-main-container table{width:100%;background-color:var(--ui-background-01);border-collapse:collapse;border:2px solid var(--ui-background-09);font-family:'Lato',sans-serif;color:var(--ui-background-09)}#snowflake-blog-template-main-container table thead{background-color:var(--ui-01)}#snowflake-blog-template-main-container table th,#snowflake-blog-template-main-container table td{border:2px solid var(--ui-background-09);padding:var(--spacing-01)}",":type":"snowflake-site/components/markup-editor","isGSAPEnabled":false},"experiencefragment-footer":{"id":"experiencefragment-35d4ce33ed","localizedFragmentVariationPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/footer/master/jcr:content","configured":true,":type":"snowflake-site/components/experiencefragment","xfModelPath":"/content/experience-fragments/snowflake-site/language-masters/en/site/footer/master.xfmodel.json"}},":itemsOrder":["experiencefragment-banner","experiencefragment-header","experiencefragment-sub-header","responsivegrid","container_47873732","experiencefragment-pre-footer","markup_editor","markup_editor-table","experiencefragment-footer"],":type":"wcm/foundation/components/responsivegrid"}},":itemsOrder":["root"],":hierarchyType":"page",":path":"/content/snowflake-site/global/en/blog/engineering/snowflake-coco-cowork-token-spend-efficiency","locale":"en"}