Kimi K3 vs Qwen3.8 27B
Wins 5 of 8 areas
Coding · Agents · Reasoning · Images and charts · Long documents
Wins 2 of 8 areas
Math · Following instructions
Kimi K3 is the stronger all-rounder.Qwen3.8 27B is cheaper and better at math.
Scores updated · 48 tests both models report · How we compare
Where each one wins
Tests won in each of the eight areas we test. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- ReasoningHard problems that need careful thinking50Kimi K35 of 6 tests · 1 tie
- CodingWriting and fixing software40Kimi K34 of 5 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own41Kimi K34 of 6 tests · 1 tie
- Images and chartsUnderstanding pictures, charts and video20Kimi K32 of 2 tests
- Long documentsFinding answers in very long texts20Kimi K32 of 2 tests
- MathCompetition and research-level math01Qwen3.8 27B1 of 1 test
- Following instructionsDoing exactly what it is asked01Qwen3.8 27B1 of 1 test
- FactsGetting facts right instead of making them up11Even1 each
Math and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Qwen3.8 27B costs 81% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Kimi K3 pulls ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+32points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+29.6points ahead
- Abstract visual puzzles that people can solveARC-AGI-2+18points ahead
Where Qwen3.8 27B pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+22.9points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+2points ahead
- Fresh competition math problemsLiveBench · Mathematics+1.8points ahead
Every test, side by side
All 48 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K3
- SciCodeKimi K3 by 12.959.546.6+12.9
- LMArena · WebDevKimi K3 by 69 rating points16591590+69 rating
- LiveBench · CodingKimi K3 by 5.881.575.7+5.8
- Terminal-Bench 2.1Kimi K3 by 5.28579.8+5.2
- LiveBench · Agentic Codingtie62.261.4tie
AgentsKimi K3
- BrowseCompKimi K3 by 29.691.261.6+29.6
- Terminal-Bench 4.0Kimi K3 by 712.65.6+7
- MCP AtlasKimi K3 by 6.984.277.3+6.9
- GDPValKimi K3 by 5.751.746+5.7
- τ-Bench V3 · BankingQwen3.8 27B by 24648+2
- OSWorld-Verifiedtie84.884.3tie
ReasoningKimi K3
- ARC-AGI-2Kimi K3 by 1860.442.4+18
- CritPtKimi K3 by 1823.45.4+18
- Humanity's Last ExamKimi K3 by 1346.933.9+13
- LiveBench · ReasoningKimi K3 by 10.790.780+10.7
- GPQA DiamondKimi K3 by 393.590.5+3
- SimpleBenchtie60.760.2tie
FactsEven
- AA-Omniscience · AccuracyKimi K3 by 3247.615.6+32
- AA-Omniscience · Non-hallucinationQwen3.8 27B by 22.946.869.7+22.9
Images and chartsKimi K3
- MMMU-ProKimi K3 by 4.280.576.3+4.2
- CharXiv (RQ)Kimi K3 by 1.191.390.2+1.1
Long documentsKimi K3
- LongBench v2Kimi K3 by 16.76952.3+16.7
- AA-LCRKimi K3 by 6.788.782+6.7
Following instructionsQwen3.8 27B
- LiveBench · Instruction FollowingQwen3.8 27B by 1.371.472.7+1.3
Other results23 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- CritPt v2 HardKimi K3 by 36.540.54+36.5
- AA-OmniscienceKimi K3 by 29.719.7-10+29.7
- DeepSWE 1.1Kimi K3 by 26.86942.2+26.8
- JobBenchKimi K3 by 19.552.933.4+19.5
- NL2RepoKimi K3 by 15.75842.3+15.7
- DeepSearchQA (F1)Kimi K3 by 15.39579.7+15.3
- Agents' Last ExamQwen3.8 27B by 14.628.342.9+14.6
- Chartography originalKimi K3 by 14.426.612.2+14.4
- SWEBench Pro v2 (Hard)Kimi K3 by 14.488.273.8+14.4
- GDPval-AA 2.1Kimi K3 by 116 rating points15431427+116 rating
- HLE-DiamondKimi K3 by 11.626.915.3+11.6
- livebench_languageKimi K3 by 11.185.574.3+11.1
- AA IntelligenceKimi K3 by 9.943.633.7+9.9
- Toolathlon VerifiedKimi K3 by 9.476.567.1+9.4
- Artificial Analysis Coding IndexKimi K3 by 8.176.268.1+8.1
- Terminal-Bench-Science 0.1Kimi K3 by 8.18.10+8.1
- ARC-AGI-1Kimi K3 by 794.587.5+7
- Terminal-Bench 4.0Kimi K3 by 712.65.6+7
- GDP (Surge AI)Kimi K3 by 5.62216.4+5.6
- AA Agentic IndexKimi K3 by 4.150.646.5+4.1
- MATH-VisionKimi K3 by 3.297.894.6+3.2
- livebench_data_analysisKimi K3 by 2.178.776.6+2.1
- BabyVisiontie85.785.6tie
Questions people ask
Which is better, Kimi K3 or Qwen3.8 27B?
Kimi K3 wins five of the eight areas we test: coding, agents, reasoning, images and charts and long documents. Qwen3.8 27B wins math and following instructions, and costs 81% less. They are level on facts.
Which is better for coding?
Kimi K3. It wins 4 of the 5 coding tests both models report; Qwen3.8 27B wins none, and 1 is a tie.
Which is cheaper?
Kimi K3 costs $3.00 per million input tokens and $15.00 per million output tokens; Qwen3.8 27B costs $0.50 and $3.00. That makes Qwen3.8 27B about 81% cheaper for the same work.
How do you compare the two?
We use the 48 benchmark tests both models have published scores on. The verdict counts the 25 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 23 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.