--- title: DeepSeek V4 Flash 0731 | daily.dev description: DeepSeek V4 Flash 0731 has been evaluated on the ARC-AGI benchmarks. At max effort, it scores 89.0% on ARC-AGI-1 Semi-Private at $0.02 per task and 61.4% on... canonical: https://daily.dev/posts/deepseek-v4-flash-0731-ugduwmpup twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: DeepSeek V4 Flash 0731 | daily.dev og:description: DeepSeek V4 Flash 0731 has been evaluated on the ARC-AGI benchmarks. At max effort, it scores 89.0% on ARC-AGI-1 Semi-Private at $0.02 per task and 61.4% on... og:url: https://daily.dev/posts/deepseek-v4-flash-0731-ugduwmpup og:image: https://api.daily.dev/og/posts/UGduwMPUP.png og:image:alt: DeepSeek V4 Flash 0731 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 **[Hacker News](https://daily.dev/sources/hn)** · 9 min read · 0 upvotes · 0 comments ## Summary DeepSeek V4 Flash 0731 has been evaluated on the ARC-AGI benchmarks. At max effort, it scores 89.0% on ARC-AGI-1 Semi-Private at $0.02 per task and 61.4% on ARC-AGI-2 Semi-Private at $0.04 per task. The page presents verified scores across three reasoning effort levels (Max, High, Low) and per-task pass/fail breakdowns for both ARC-AGI-1 (400 tasks) and ARC-AGI-2 (120 tasks) public evaluation sets. ARC-AGI-3 results are not yet available for this model. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What does DeepSeek V4 Flash 0731 score on ARC-AGI-2? DeepSeek V4 Flash 0731 scores 61.4% on ARC-AGI-2 Semi-Private at max effort, costing $0.04 per task. At high effort it scores 56.0%, and at low effort 46.0%. The model has not yet been evaluated on ARC-AGI-3. _Developers tracking AI reasoning model progress on ARC-AGI follow benchmark updates like these on daily.dev._ ### How much does it cost to run DeepSeek V4 Flash 0731 on ARC-AGI-1 at max effort? Running DeepSeek V4 Flash 0731 on ARC-AGI-1 at max effort costs $0.02 per task, achieving 89.0% on the Semi-Private evaluation. At high effort the score drops to 87.0%, and at low effort to 84.0%. _Teams choosing between reasoning models for cost-sensitive workloads track pricing and performance tradeoffs on daily.dev._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 141 comments across hackernews (as of 2026-08-07). **TL;DR:** Commenters are broadly impressed by DeepSeek V4 Flash 0731's benchmark scores and extremely low cost, with many sharing real-world usage stories; the main concern is DeepSeek's announced upcoming price increase, though the open-weight nature of the model means third-party providers can keep prices competitive. **Sentiment:** 62% positive · 28% mixed · 10% skeptical **The case for** - The model's cost is so low — especially for cached tokens — that many previously impractical use cases (CI auto-fixes, continuous log monitoring, social feed filtering) become economically viable. - Cache hit rates of 95–98%+ in coding sessions make the effective per-task cost dramatically lower than competitors, even at listed token prices. - Real-world users report quality that punches well above its price tier, with some preferring it over Claude for programming due to different and complementary blind spots. - Being open-weight with an MIT license means third-party providers can serve it competitively even if DeepSeek raises its own API prices. - Running a 2-bit quant locally on consumer hardware (128 GB Strix Halo) is reportedly feasible and still yields good results. **The pushback** - DeepSeek has officially warned of a 'significant' upcoming price increase on its own API, which could shift the cost-performance picture. - Third-party providers charge roughly 10x more for cached input tokens than DeepSeek itself, eroding the headline cost advantage for heavy users. - Some users report the model is slower than the preview version and token-verbose, making latency a concern for interactive use. - A minority of users found the model unhelpful for their tasks, and OpenRouter's serving quality (cache utilization, max-effort support) was flagged as inconsistent. - Auto-generating tests in CI was called out as a bad default use case without careful prompting and context. **By community** - hackernews (positive): Developers are enthusiastic about the price-to-performance ratio and share concrete usage data, while debating the implications of the announced price hike and the uniqueness of DeepSeek's caching advantage. **Hottest debate:** Whether DeepSeek's upcoming price increase will meaningfully erode its cost advantage given that open-weight third-party providers can undercut it — except on cached token pricing, which remains uniquely cheap. **Open questions** - Will other inference providers replicate DeepSeek's KV-cache efficiency (e.g., via vLLM disk-offload PRs) to match its cache pricing? - Why haven't major US labs adopted Multi-head Latent Attention (MLA) despite its ~2-year availability and clear caching benefits? - When will DeepSeek V4 Pro be released, and will it maintain the same pricing model? - What are the ARC-AGI-3 results for this model? **Highlights** > My usage thus far from api.deepseek.com - input_cache_hit_tokens: 1,265,646,976 x 0.0000000028 = $3.5438115328 - input_cache_miss_tokens: 18,208,088 x 0.00000014 = $2.54913232 - output_tokens: 9,615,178 x 0.00000028 = $2.69224984 - request_count: 10,837 (no price) Total cost: $8.7851936928 (approximately $8.79) Cache: - Hit: 1,265,646,976 - Miss: 18,208,088 - Total input tokens: 1,283,855,064 Hit rate: 98.582% (1,265,646,976 / 1,283,855,064) > — [greenavocado on hackernews](https://news.ycombinator.com/item?id=49215348) > There's more to inference than just the input/output token cost. Caching has a massive impact. Deepseek charges $0.0028 per cache read on Openrouter. The next cheapest is $0.018. That's a massive difference and quickly adds up on coding sessions (which often hit 95%+ cached tokens). > — [LaurensBER on hackernews](https://news.ycombinator.com/item?id=49214650) > > Deepseeks secret sauce is the incredibly cheap caching (magnitude cheaper than other providers). Can anyone working at one of the main US labs (Google, OpenAI, Anthropic) comment on WTF they haven't even tried MLA - despite the obvious massive advantages? I know enough to know they aren't completely incompetent. So there must be a quite good reason. But it remains a mystery to me. DeepSeek's MLA is like almost 2 years old at this time. They've got thousands of people working on this stuff. They clearly have the ability to at least try it... > — [onlyrealcuzzo on hackernews · 2 comments](https://news.ycombinator.com/item?id=49214722) > I've posted a few times about my project that's a collection of 30k-250k webapps that are served from a WebDAV server. The apps know how to write updated copies of themselves back to the server. My family uses it. I have gallery apps (yearbooks for each year are a lot of fun!) of us on trips and just living, an outlining app that's a mesh of Workflowy and Org Mode (it's called Fluxtral), a markdown-backed app (it uses marked.min.js, and is called Dextral) that offers documents, logs, calendars, and kanban boards, all parsed from markdown. I have a List app for gear, trips, shopping, etc. that we all can contribute to. There are utilities (world clock, calendar) and games (an oracle for RPGs, a KenKen implementation), and apps (a diagram editor that exports to SVG, a web-launcher that uses pneumonics, a Scheme-based hacking environment, and a spreadsheet that does most of what you'd expect aside from Solver and Pivot tables). I started these projects before AI, and made slow progress over the years, but the modern versions of all this stuff have been built with Deepseek V4 Flash. I've also used Gemini in the very early days, and Kimi K2.6 later on, but these days, since I can now host Deepseek v4 Flash 0731 in a 2-bit quant on my Strix Halo box (128GB, but only about 250GB/s of memory bandwidth, so 15t/s), I used Deepseek with omp for almost everything. It's a very capable model, and I'm amazed I can run it locally and get good results. It's really revolutionary for my (small) use cases. > — [rpdillon on hackernews](https://news.ycombinator.com/item?id=49215429) > I strongly recommend trying this for programming tasks. It is strong (not Fable strong though) with a much better “persona” than Opus, and very different blindspots. If you flip between Claude and this you will find both catch the mistakes of the other before they get out of control. On balance I actually prefer DeepSeek for programming now, because of the way it talks. > — [mosura on hackernews · 2 comments](https://news.ycombinator.com/item?id=49214652) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=49214008) · 182 points · 141 comments ## Similar posts on daily.dev - [Intelligence, Performance & Price Analysis](https://daily.dev/posts/intelligence-performance-price-analysis-uowbhxnia) · Hacker News · 10 upvotes · 0 comments - [DeepSeek-V4.1-Flash on vLLM: 5x Agentic Throughput Since Day 0](https://daily.dev/posts/deepseek-v4-1-flash-on-vllm-5x-agentic-throughput-since-day-0-js5lna7tm) · vLLM · 0 upvotes · 0 comments - [DeepSeek V4 Flash now runs updated weights on AI Gateway](https://daily.dev/posts/deepseek-v4-flash-now-runs-updated-weights-on-ai-gateway-9eqhpaore) · Vercel · 2 upvotes · 0 comments --- Tags: [#ai](https://daily.dev/tags/ai), [#llm](https://daily.dev/tags/llm), [#deepseek](https://daily.dev/tags/deepseek) [View this post on daily.dev](https://daily.dev/posts/deepseek-v4-flash-0731-ugduwmpup) ```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","url":"https://daily.dev/posts/deepseek-v4-flash-0731-ugduwmpup","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/deepseek-v4-flash-0731-ugduwmpup"},"datePublished":"2026-08-07T18:35:26.839Z","dateModified":"2026-08-07T20:34:23.064Z","description":"DeepSeek V4 Flash 0731 has been evaluated on the ARC-AGI benchmarks. At max effort, it scores 89.0% on ARC-AGI-1 Semi-Private at $0.02 per task and 61.4% on...","image":"https://media.daily.dev/image/upload/s--HRgLpUt6--/f_auto/v1722860399/public/Placeholder%2003","thumbnailUrl":"https://media.daily.dev/image/upload/s--HRgLpUt6--/f_auto/v1722860399/public/Placeholder%2003","isAccessibleForFree":true,"articleSection":"Hacker News","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":"Hacker News","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/hn","url":"https://daily.dev/sources/hn"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/deepseek-v4-flash-0731-ugduwmpup","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai,llm,deepseek","timeRequired":"PT9M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Hacker News","item":"https://daily.dev/sources/hn"},{"@type":"ListItem","position":3,"name":"DeepSeek V4 Flash 0731"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/deepseek-v4-flash-0731-ugduwmpup#faq","mainEntity":[{"@type":"Question","name":"What does DeepSeek V4 Flash 0731 score on ARC-AGI-2?","acceptedAnswer":{"@type":"Answer","text":"DeepSeek V4 Flash 0731 scores 61.4% on ARC-AGI-2 Semi-Private at max effort, costing $0.04 per task. At high effort it scores 56.0%, and at low effort 46.0%. The model has not yet been evaluated on ARC-AGI-3. Developers tracking AI reasoning model progress on ARC-AGI follow benchmark updates like these on daily.dev."}},{"@type":"Question","name":"How much does it cost to run DeepSeek V4 Flash 0731 on ARC-AGI-1 at max effort?","acceptedAnswer":{"@type":"Answer","text":"Running DeepSeek V4 Flash 0731 on ARC-AGI-1 at max effort costs $0.02 per task, achieving 89.0% on the Semi-Private evaluation. At high effort the score drops to 87.0%, and at low effort to 84.0%. Teams choosing between reasoning models for cost-sensitive workloads track pricing and performance tradeoffs on daily.dev."}}]} ```