--- title: Kimi K3.1 Pro and K3.1 Pro Highspeed | daily.dev description: Kimi Code now offers six model IDs, led by the newly rolled-out K3.1 Pro, a 3T-parameter flagship with a 1M context window and thinking-mode-only operation. A... canonical: https://daily.dev/posts/kimi-k3-1-pro-and-k3-1-pro-highspeed-jqqnhyupr twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Kimi K3.1 Pro and K3.1 Pro Highspeed | daily.dev og:description: Kimi Code now offers six model IDs, led by the newly rolled-out K3.1 Pro, a 3T-parameter flagship with a 1M context window and thinking-mode-only operation. A... og:url: https://daily.dev/posts/kimi-k3-1-pro-and-k3-1-pro-highspeed-jqqnhyupr og:image: https://api.daily.dev/og/posts/jQQNHYuPR.png og:image:alt: Kimi K3.1 Pro and K3.1 Pro Highspeed 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. # Kimi K3.1 Pro and K3.1 Pro Highspeed **[Hacker News](https://daily.dev/sources/hn)** · 8 min read · 0 upvotes · 0 comments ## Summary Kimi Code now offers six model IDs, led by the newly rolled-out K3.1 Pro, a 3T-parameter flagship with a 1M context window and thinking-mode-only operation. A Highspeed variant runs 2 to 3 times faster with identical capabilities. Other options include K3 (1M or 256K context), K2.8 Preview, and K2.7 Code Highspeed. Switching models or thinking effort mid-session invalidates the context cache and increases token usage, and a mistyped model ID falls back to K2.8 Preview without thinking. The API supports OpenAI Responses/Chat Completions and Anthropic protocols for third-party tools like Claude Code, OpenCode, and Codex. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What model ID do I use for Kimi K3.1 Pro Highspeed in a third-party coding tool? Use the model ID k3.1-pro-highspeed. It has the same capabilities as k3.1-pro but delivers roughly 2 to 3 times the speed, at about 2 times the quota usage. Entering a model version name instead of this exact ID, or mistyping it, causes requests to silently fall back to K2.8 Preview without thinking. _daily.dev helps developers wiring up coding agents catch config details like exact model IDs before they ship._ ### Why is my Kimi K3.1 Pro API request returning HTTP 400 with thinking disabled? K3.1 Pro models only support thinking mode, so any request with thinking set to off, none, or disabled is rejected with HTTP 400. This differs from the K3 series, K2.8 Preview, and K2.7 Code Highspeed, which instead get routed to K2.8 Preview without thinking rather than erroring out. _Developers debugging LLM API errors can track model-specific quirks like this one on daily.dev._ ### Does switching from K3 1M context to K3-256k lose my conversation history? It can, if the session already exceeds 256k tokens, because tools like Kimi Code CLI and Claude Code will perform a compact on the tool side first. The recommended approach is to manually run compact before switching to preserve key task context, and to compact first if the history includes video, since K3-256k does not support video input. _daily.dev keeps developers managing long-context LLM sessions current on context-limit gotchas like this._ ## Community take How the wider developer community reacted, aggregated from 1 discussion and 143 comments across hackernews (as of 2026-10-23). **TL;DR:** Benchmarks (notably a 55 on Artificial Analysis, 4th overall) impressed many, with pricing seen as competitive with Opus 5.5 and Sol, but much of the discussion focused on Moonshot's deliberate removal of cyber-offense capability from the open-weight release, with people split on whether that's prudent safety or a hollow gesture. **Sentiment:** 35% positive · 40% mixed · 25% skeptical **The case for** - Benchmark scores and pricing look competitive with Opus 5.5 and other frontier models - The methodology for removing cyber-related training data is seen by some as a reasonable, lightweight, and verifiable approach - Reported price drop versus the previous K3 model despite being larger - Faster response speed versus K3, addressing a known complaint about slow, overly verbose thinking **The pushback** - Skepticism that nerfing cyber capability is just a cop-out or rules-for-thee-not-for-me double standard - Concern that overly cautious guardrails make models less useful for legitimate security/defensive work - Past Kimi models were seen as underperforming relative to benchmarks and expensive due to excessive reasoning token usage - Complaints about unusably restrictive subscription usage quotas (hitting limits within under an hour) - Doubts about whether 'frontier' benchmark claims like Terminal-Bench scores reflect real-world performance, citing flawed grading **By community** - hackernews (mixed): Reactions split between genuine excitement over benchmarks/pricing and a long, contentious debate over the model's deliberate removal of cyber-offense capability, alongside complaints about past Kimi reliability, cost, and usage limits. **Hottest debate:** Whether Moonshot's removal of cyber-offense training data from the open-weight release is a responsible safety measure or a disingenuous half-measure that doesn't really restrict misuse. **Open questions** - How does real-world agentic coding performance compare to the benchmark scores, especially given doubts about benchmark validity (e.g., Terminal-Bench grading issues)? - What is the actual decode tokens/sec speed for the Highspeed variant? - Will removed cyber capability be trivially reintroduced via fine-tuning once weights are released? - How usable will the subscription plans be given reported quota limits hit within under an hour? **Highlights** > > So they admit that they could do more, but they won't? Yeah, there's another paragraph: > Identifying and removing cyber-related data. During development, we observed vulnerability research performance improving steadily as post-training scaled. Tracing the source of this trend, we found that cyber-related content was embedded within the general mid-training corpus, and cyber-related tasks were present in the RL environment mix — a natural byproduct of training on broad data. In preliminary experiments on internal models, we found that without data removal at either the mid-training or the RL stage, downstream cyber capability would increase substantially. These findings gave us a concrete lever for control. For the open-weight release of K3.1 Pro, we identified and removed cyber-related data from both training stages, while retaining the general-purpose coding tasks needed for model usefulness. We will continue to apply appropriate controls at every training stage, of which cyber-related data identification and restriction is one example. (emphasis mine) It's clear that cyber is a touchy subject right now. But it's also clear that RL works on it, because it's highly verifiable. So smaller models are going to become very interesting, because they can be fine-tuned later to cover this field. > — [NitpickLawyer on hackernews · 1 comments](https://news.ycombinator.com/item?id=50212851) > Extrapolating a bit, Moonshot AI provides Kimi K3 at $15 / $3 / $0.3 per million output/input/cached tokens. A 30% reduction would put K3.1 at roughly $12 / $2.4 / $0.24. For comparison: Sol 6.1: $10 / $2 / $0.10 Opus 5.5: $20 / $4 / $0.20 Astra 6: $50 / $10 / $1 Fable 5.1: $50 / $10 / $0.25 So Kimi K3.1 Pro would be generally competitive with the latest Sol and Opus models on API pricing. And that's before factoring in third-party providers, some of which serve Kimi K3 for significantly less than Moonshot. For example, inference.net offers K3 for $10.80 / $0.239 / $0.21. (One caveat: A Claude or GPT subscription still gives you way more value compared to API pricing.) Exciting times in any case! We really need more direct competition to OpenAI and Anthropic. --- Edit: As another comment pointed out, official pricing is now available and it's about 2/3 the price of Kimi K3: https://platform.kimi.ai/docs/pricing/chat#model-pricing > — [RussianCow on hackernews](https://news.ycombinator.com/item?id=50212660) > To be fair, some of the benchmarks they highlight might not represent the reality, for example they say: > On Terminal-Bench 4.0, which evaluates coding agents on realistic, resource-constrained terminal tasks using strict task-specific tests, K3.1 Pro scores 65.7%, ahead of GPT-6 Astra (58.2%) and narrowly behind Claude Opus 5.5 (66.4%). Now see https://preferencemodel.com/blog/terminal-bench-4-is-saturat...: > Our biggest finding is that Terminal Bench 4 is closer to solved than its leaderboard suggests. In 41 of the 65 tasks we reviewed, models could do honest, correct work and still fail, because the grader checked a rule the instructions never clearly stated. We also caught frontier models cheating their way to full marks. > With the same settings as on Harbor, Claude Sonnet 5.5 went from passing 62.8% of its attempts on these 65 tasks to 89.2%. The best score on Harbor's leaderboard for the same tasks is 65.8%, by Claude Opus 5.5. > — [Tiberium on hackernews](https://news.ycombinator.com/item?id=50212575) > I bought a yearly subscription to kimi when it was the hype and never use it. The 5h limit makes it completely unusable. You barely start doing anything and you hit the limit immediately. I advise everyone against using kimi. > — [Athanase000 on hackernews](https://news.ycombinator.com/item?id=50212857) **Source threads** - [hackernews](https://news.ycombinator.com/item?id=50212065) · 90 points · 143 comments ## Similar posts on daily.dev - [Kimi K3-256k](https://daily.dev/posts/kimi-k3-256k-tk0yt6bbf) · Hacker News · 0 upvotes · 0 comments - [Moonshot AI’s Kimi K2.7-Code Targets Token Efficiency in Agentic Coding](https://daily.dev/posts/moonshot-ai-s-kimi-k2-7-code-targets-token-efficiency-in-agentic-coding-rpll6qhkm) · DevOps.com · 0 upvotes · 0 comments --- Tags: [#llm](https://daily.dev/tags/llm), [#claude-code](https://daily.dev/tags/claude-code) [View this post on daily.dev](https://daily.dev/posts/kimi-k3-1-pro-and-k3-1-pro-highspeed-jqqnhyupr) ```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":"Kimi K3.1 Pro and K3.1 Pro Highspeed","url":"https://daily.dev/posts/kimi-k3-1-pro-and-k3-1-pro-highspeed-jqqnhyupr","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/kimi-k3-1-pro-and-k3-1-pro-highspeed-jqqnhyupr"},"datePublished":"2026-10-23T17:24:24.353Z","dateModified":"2026-10-23T18:53:07.548Z","description":"Kimi Code now offers six model IDs, led by the newly rolled-out K3.1 Pro, a 3T-parameter flagship with a 1M context window and thinking-mode-only operation. A...","image":"https://media.daily.dev/image/upload/s--P4t4XyoV--/f_auto/v1722860399/public/Placeholder%2001","thumbnailUrl":"https://media.daily.dev/image/upload/s--P4t4XyoV--/f_auto/v1722860399/public/Placeholder%2001","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/kimi-k3-1-pro-and-k3-1-pro-highspeed-jqqnhyupr","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"llm,claude-code","timeRequired":"PT8M"} {"@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":"Kimi K3.1 Pro and K3.1 Pro Highspeed"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/kimi-k3-1-pro-and-k3-1-pro-highspeed-jqqnhyupr#faq","mainEntity":[{"@type":"Question","name":"What model ID do I use for Kimi K3.1 Pro Highspeed in a third-party coding tool?","acceptedAnswer":{"@type":"Answer","text":"Use the model ID k3.1-pro-highspeed. It has the same capabilities as k3.1-pro but delivers roughly 2 to 3 times the speed, at about 2 times the quota usage. Entering a model version name instead of this exact ID, or mistyping it, causes requests to silently fall back to K2.8 Preview without thinking. daily.dev helps developers wiring up coding agents catch config details like exact model IDs before they ship."}},{"@type":"Question","name":"Why is my Kimi K3.1 Pro API request returning HTTP 400 with thinking disabled?","acceptedAnswer":{"@type":"Answer","text":"K3.1 Pro models only support thinking mode, so any request with thinking set to off, none, or disabled is rejected with HTTP 400. This differs from the K3 series, K2.8 Preview, and K2.7 Code Highspeed, which instead get routed to K2.8 Preview without thinking rather than erroring out. Developers debugging LLM API errors can track model-specific quirks like this one on daily.dev."}},{"@type":"Question","name":"Does switching from K3 1M context to K3-256k lose my conversation history?","acceptedAnswer":{"@type":"Answer","text":"It can, if the session already exceeds 256k tokens, because tools like Kimi Code CLI and Claude Code will perform a compact on the tool side first. The recommended approach is to manually run compact before switching to preserve key task context, and to compact first if the history includes video, since K3-256k does not support video input. daily.dev keeps developers managing long-context LLM sessions current on context-limit gotchas like this."}}]} ```