--- title: Deepseek V4 Flash (0731 - Fully Tested): TOP 5 in my... description: DeepSeek V4 Flash 0731 has launched in public beta via the official API, bringing massive agentic capability improvements through post-training alone — same... canonical: https://daily.dev/posts/deepseek-v4-flash-0731---fully-tested-top-5-in-my-tests-this-is-an-actual-comeback--touk6g9qw twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Deepseek V4 Flash (0731 - Fully Tested): TOP 5 in my TESTS! This is AN ACTUAL COMEBACK!!! | daily.dev og:description: DeepSeek V4 Flash 0731 has launched in public beta via the official API, bringing massive agentic capability improvements through post-training alone — same... og:url: https://daily.dev/posts/deepseek-v4-flash-0731---fully-tested-top-5-in-my-tests-this-is-an-actual-comeback--touk6g9qw og:image: https://api.daily.dev/og/posts/touk6G9qw.png og:image:alt: Deepseek V4 Flash (0731 - Fully Tested): TOP 5 in my TESTS! This is AN ACTUAL COMEBACK!!! og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Deepseek V4 Flash (0731 - Fully Tested): TOP 5 in my TESTS! This is AN ACTUAL COMEBACK!!! **[AICodeKing](https://daily.dev/sources/aicodeking)** · 11 min read · 1 upvotes · 0 comments ## Summary DeepSeek V4 Flash 0731 has launched in public beta via the official API, bringing massive agentic capability improvements through post-training alone — same 284B parameter architecture, but dramatically better benchmark scores. On agentic coding benchmarks like Terminal Bench 2.1 (82.7 vs 61.8 for preview) and Deep Sway (54.4 vs 7.3), it now surpasses the V4 Pro preview and competes near Claude Opus 4.8 at a fraction of the cost. In independent KingBench testing, it scored 72.5% (58/80), setting an all-time record on the hardest 3D wristwatch task and acing a long-horizon agentic fine-tuning task. It natively supports the responses API for easy integration into coding agent workflows. The full V4 Pro release is reportedly coming soon. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What changed in the Deepseek V4 Flash 0731 update compared to the preview version? Deepseek V4 Flash 0731 keeps the exact same 284 billion total parameter mixture-of-experts architecture and 1 million token context window as the preview, but received a heavy round of agentic post-training. This pushed Terminal Bench 2.1 from 61.8 to 82.7 and Deep Sway from 7.3 to 54.4, with no increase in model size. _Developers picking a cost-effective coding model can track how Deepseek's post-training gains compare on daily.dev._ ### Does Deepseek V4 Flash support the codex-style responses API format? Yes, the official V4 Flash 0731 release natively supports the responses API format and is explicitly adapted for codex-style harnesses, similar to adaptations GLM and other Chinese labs have made. This is intended to let it plug into codex-style agentic coding workflows without significant friction. _Anyone wiring agentic coding tools together can follow harness-compatibility updates like this on daily.dev._ ### How does Deepseek V4 Flash 0731 compare to Opus 4.8 and GLM 5.2 on agentic coding benchmarks? On official benchmarks, V4 Flash 0731 scores 82.7 on Terminal Bench 2.1, above GLM 5.2's 81 and close to Opus 4.8's 85, despite being a much smaller and cheaper model. On an independent coding benchmark it scored 72.5%, landing just below GLM 5.2's 75% and just above GPT 5.6 Saul's 71.25%. _Developers weighing price versus performance across coding models can compare results like these via daily.dev._ ## Similar posts on daily.dev - [DeepSeek API Docs](https://daily.dev/posts/deepseek-api-docs-mqussratb) · Hacker News · 0 upvotes · 0 comments - [OpenCode \+ DeepSeek V4: The Budget Combo That Doesn't Feel Like a Budget Combo](https://daily.dev/posts/opencode-deepseek-v4-the-budget-combo-that-doesn-t-feel-like-a-budget-combo-d9ybqvwwf) · Playful Programming · 3 upvotes · 0 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#llm](https://daily.dev/tags/llm), [#ai-coding](https://daily.dev/tags/ai-coding), [#agentic-ai](https://daily.dev/tags/agentic-ai), [#deepseek](https://daily.dev/tags/deepseek) [View this post on daily.dev](https://daily.dev/posts/deepseek-v4-flash-0731---fully-tested-top-5-in-my-tests-this-is-an-actual-comeback--touk6g9qw) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"Deepseek V4 Flash (0731 - Fully Tested): TOP 5 in my TESTS! This is AN ACTUAL COMEBACK!!!","url":"https://daily.dev/posts/deepseek-v4-flash-0731---fully-tested-top-5-in-my-tests-this-is-an-actual-comeback--touk6g9qw","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/deepseek-v4-flash-0731---fully-tested-top-5-in-my-tests-this-is-an-actual-comeback--touk6g9qw"},"datePublished":"2026-07-31T11:27:27.181Z","dateModified":"2026-09-14T08:04:39.378Z","description":"DeepSeek V4 Flash 0731 has launched in public beta via the official API, bringing massive agentic capability improvements through post-training alone — same...","image":"https://i.ytimg.com/vi/weRPJ1YzWvY/sddefault.jpg","thumbnailUrl":"https://i.ytimg.com/vi/weRPJ1YzWvY/sddefault.jpg","isAccessibleForFree":true,"articleSection":"AICodeKing","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"AICodeKing","logo":"https://media.daily.dev/image/upload/s--x7nDUfWj--/f_auto,q_auto/v1768208312/logos/aicodeking","url":"https://daily.dev/sources/aicodeking"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/deepseek-v4-flash-0731---fully-tested-top-5-in-my-tests-this-is-an-actual-comeback--touk6g9qw","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai,llm,ai-coding,agentic-ai,deepseek","timeRequired":"PT11M","video":{"@type":"VideoObject","name":"Deepseek V4 Flash (0731 - Fully Tested): TOP 5 in my TESTS! This is AN ACTUAL COMEBACK!!!","description":"DeepSeek V4 Flash 0731 has launched in public beta via the official API, bringing massive agentic capability improvements through post-training alone — same...","thumbnailUrl":"https://i.ytimg.com/vi/weRPJ1YzWvY/sddefault.jpg","uploadDate":"2026-07-31T11:27:27.181Z","duration":"PT11M","url":"https://api.daily.dev/r/touk6G9qw","embedUrl":"https://www.youtube.com/embed/weRPJ1YzWvY"}} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"AICodeKing","item":"https://daily.dev/sources/aicodeking"},{"@type":"ListItem","position":3,"name":"Deepseek V4 Flash (0731 - Fully Tested): TOP 5 in my TESTS! This is AN ACTUAL COMEBACK!!!"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/deepseek-v4-flash-0731---fully-tested-top-5-in-my-tests-this-is-an-actual-comeback--touk6g9qw#faq","mainEntity":[{"@type":"Question","name":"What changed in the Deepseek V4 Flash 0731 update compared to the preview version?","acceptedAnswer":{"@type":"Answer","text":"Deepseek V4 Flash 0731 keeps the exact same 284 billion total parameter mixture-of-experts architecture and 1 million token context window as the preview, but received a heavy round of agentic post-training. This pushed Terminal Bench 2.1 from 61.8 to 82.7 and Deep Sway from 7.3 to 54.4, with no increase in model size. Developers picking a cost-effective coding model can track how Deepseek's post-training gains compare on daily.dev."}},{"@type":"Question","name":"Does Deepseek V4 Flash support the codex-style responses API format?","acceptedAnswer":{"@type":"Answer","text":"Yes, the official V4 Flash 0731 release natively supports the responses API format and is explicitly adapted for codex-style harnesses, similar to adaptations GLM and other Chinese labs have made. This is intended to let it plug into codex-style agentic coding workflows without significant friction. Anyone wiring agentic coding tools together can follow harness-compatibility updates like this on daily.dev."}},{"@type":"Question","name":"How does Deepseek V4 Flash 0731 compare to Opus 4.8 and GLM 5.2 on agentic coding benchmarks?","acceptedAnswer":{"@type":"Answer","text":"On official benchmarks, V4 Flash 0731 scores 82.7 on Terminal Bench 2.1, above GLM 5.2's 81 and close to Opus 4.8's 85, despite being a much smaller and cheaper model. On an independent coding benchmark it scored 72.5%, landing just below GLM 5.2's 75% and just above GPT 5.6 Saul's 71.25%. Developers weighing price versus performance across coding models can compare results like these via daily.dev."}}]} ```