Kimi K3vsQwen3.8-27B
Kimi K3 | Qwen3.8-27B | |
|---|---|---|
| Specifications | ||
Parameters | 2.8T | 27B |
Context window | 1M | 262k |
| Benchmarks | ||
Nonsense detection BullshitBench v2Given a confidently-worded but nonsensical prompt, does the AI spot that it makes no sense and push back — instead of playing along and inventing an answer? The score is how often it clearly called out the nonsense. Higher is better. | 73% | — |
Agentic coding SWE-Bench ProCan the AI fix real bugs in real software? It's handed actual problems from open-source projects and has to write code that genuinely solves them. Higher is better. | — | 61.7% |
Agentic coding DeepSWE 1.1Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. (Version 1.1 of the test.) Higher is better. | 69%Best | 42.2% |
Agentic coding DeepSWE 1.0Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. Higher is better. | 67.5% | — |
Next.js coding Next.js EvalsVercel's open eval of how well AI coding agents build and migrate real Next.js apps — measured as the share of tasks the agent completes successfully. Higher is better. | 92% | — |
Supabase coding Supabase Evals · with skillsSupabase's own open benchmark: a coding agent is dropped into a real Supabase project and asked to do real work — set up a schema, fix a broken security policy, debug an Edge Function — and every run is checked against a live Supabase stack. This is the headline number, where the agent has Supabase's own skills loaded, as most people building on Supabase would. The score is the share of scenarios it got right. Higher is better. | 86.4% | — |
Supabase coding Supabase Evals · no skillsThe same Supabase scenarios, but with none of Supabase's skills loaded — so it measures what the model already knows about building on Supabase, rather than how well it follows Supabase's supplied instructions. Higher is better. | 90.9% | — |
Repo-level code generation NL2Repo-BenchTests whether the AI can turn a natural-language requirement into working code across an entire repository, not just produce a single function or patch. Higher is better. | — | 42.3% |
Software engineering QwenSWEBenchQwen's in-house coding benchmark for evaluating a model's ability to complete software-engineering work. Higher is better. | — | 79% |
Competitive coding LiveCodeBenchCoding problems published so recently the AI can't have seen them in training — a contamination-free test of raw programming skill. Higher is better. | — | 90.3% |
Agentic terminal coding Terminal-Bench 2.1Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Higher is better. | 88.3%Best | 73% |
Multi-step tool use MCP AtlasCan the AI chain together many tools and steps to complete one bigger task, rather than doing just a single thing? Higher is better. | 84.2% | — |
Professional tool use JobBenchTests the AI on professional workplace tasks that require using real work tools — the kind of multi-step jobs an office worker handles. Higher is better. | 52.9%Best | 33.4% |
Long-horizon office work CoWorkBenchTests long-running office tasks across fields including computer science, finance, law, medicine, and other productivity work. Higher is better. | — | 70.7% |
Personal tool use Toolathlon-VerifiedTests how well the AI uses everyday personal tools and apps to get things done — a human-checked version of Toolathlon. Higher is better. | 73.2% | — |
Web browsing BrowseCompCan the AI browse the web and track down hard-to-find answers? Higher is better. | 91.2% | — |
Multidisciplinary reasoning Humanity's Last Exam · no toolsHumanity's Last Exam — extremely hard expert questions across many subjects, written so you can't just look up the answer. “No tools” means the AI answers on its own. Higher is better. | 43.5%Best | 30.8% |
Multidisciplinary reasoning Humanity's Last Exam · with toolsHumanity's Last Exam — extremely hard expert questions across many subjects. “With tools” means the AI is allowed to search the web or run code while answering. Higher is better. | 56% | — |
Science GPQA DiamondGraduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better. | 93.5%Best | 89.2% |
Instruction following IFBenchTests whether the AI can follow detailed instructions and satisfy multiple constraints at once. Higher is better. | — | 79.5% |
Agentic computer use Agent's Last Exam · pass@1A hard set of desktop and operating-system tasks an AI agent has to finish by looking at the screen and working the machine itself. The score is the share it passes outright — partial credit does not count. Higher is better. | — | 20.4% |
Agentic computer use Agent's Last Exam · scoreThe graded score on the same desktop and operating-system tasks in Agent's Last Exam, giving partial credit for progress beyond the strict pass-or-fail result. Higher is better. | — | 42.9% |
Knowledge work GDPval-AA v2economically valuable knowledge work (v2, re-based Elo) | 1668 | — |
Chart reasoning CharXiv ReasoningCan the AI read and reason about complex charts and figures, not just text? Higher is better. | 84.8% | — |
Multimodal reasoning MMMU-ProA tougher version of MMMU — college-level questions that mix images, diagrams, and text together. Higher is better. | 81.6% | — |
Community preference Arena Elo (Text)Real people chat with two anonymous AIs side by side and vote for the answer they prefer. Votes become a chess-style Elo rating on arena.ai — it measures which AI people actually like, not test scores. Higher is better. | 1486 | — |
Community preference (code) Arena Elo (Code)Like the text arena, but people vote on which AI writes better code. The votes become a chess-style Elo rating on arena.ai. Higher is better. | 1679 | — |
| Overview | ||
| Company | Moonshot AI | Qwen |
| Release date | Jul 16 2026 | Aug 14 2026 |
| Access | Open Weight | Open Weight |
Which is better: Kimi K3 or Qwen3.8-27B?
Kimi K3 leads Qwen3.8-27B on 5 of the 5 benchmarks they both report. Kimi K3 shipped 29 days before Qwen3.8-27B, so benchmark comparisons should account for the intervening progress.
Kimi K3 has 2.8T parameters, while Qwen3.8-27B has 27B. Context windows are 1M (Kimi K3) vs 262k (Qwen3.8-27B).
On DeepSWE 1.1, Kimi K3 leads at 69% vs Qwen3.8-27B at 42.2%. On Terminal-Bench 2.1, Kimi K3 leads at 88.3% vs Qwen3.8-27B at 73%. On JobBench, Kimi K3 leads at 52.9% vs Qwen3.8-27B at 33.4%. On Humanity's Last Exam · no tools, Kimi K3 leads at 43.5% vs Qwen3.8-27B at 30.8%. On GPQA Diamond, Kimi K3 leads at 93.5% vs Qwen3.8-27B at 89.2%.
Frequently asked questions
Kimi K3 was released by Moonshot AI on Jul 16 2026.
Qwen3.8-27B was released by Qwen on Aug 14 2026.
Kimi K3 leads on DeepSWE 1.1 — Kimi K3 69% vs Qwen3.8-27B 42.2%.
Kimi K3 leads on Humanity's Last Exam · no tools — Kimi K3 43.5% vs Qwen3.8-27B 30.8%.
Kimi K3 has a 1M context window; Qwen3.8-27B has 262k.