Kimi K3 vs Qwen3.5 397B A17B
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents
Wins 0 of 7 areas
—
Kimi K3 is the stronger all-rounder.Qwen3.5 397B A17B is cheaper.
Scores updated · 29 tests both models report · How we compare
Where each one wins
Tests won in each of the seven areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- AgentsCarrying out multi-step tasks on its own60Kimi K36 of 6 tests
- CodingWriting and fixing software30Kimi K33 of 3 tests
- ReasoningHard problems that need careful thinking30Kimi K33 of 3 tests
- FactsGetting facts right instead of making them up30Kimi K33 of 3 tests
- MathCompetition and research-level math30Kimi K33 of 3 tests
- Images and chartsUnderstanding pictures, charts and video10Kimi K31 of 1 test
- Long documentsFinding answers in very long texts10Kimi K31 of 1 test
Images and charts and long documents 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.
Qwen3.5 397B A17B costs 77% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Kimi K3 pulls ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+41points ahead
- Real work tasks from 44 professionsGDPVal+36.9points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+35.7points ahead
Where Qwen3.5 397B A17B pulls ahead
No clear win on a test scored out of 100.
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K3
- Terminal-Bench 2.1Kimi K3 by 33.78551.3+33.7
- LMArena · WebDevKimi K3 by 255 rating points16551400+255 rating
- SciCodeKimi K3 by 14.759.544.8+14.7
AgentsKimi K3
- GDPValKimi K3 by 36.951.714.8+36.9
- τ-Bench V3 · BankingKimi K3 by 32.64613.4+32.6
- AA ApexAgentsKimi K3 by 2641.315.3+26
- BrowseCompKimi K3 by 22.291.269+22.2
- AA IT-Bench SREKimi K3 by 13.647.734.1+13.6
- Terminal-Bench 4.0Kimi K3 by 12.612.60+12.6
ReasoningKimi K3
- CritPtKimi K3 by 21.723.41.7+21.7
- Humanity's Last ExamKimi K3 by 17.946.929+17.9
- GPQA DiamondKimi K3 by 4.293.589.3+4.2
FactsKimi K3
- AA-Omniscience · Non-hallucinationKimi K3 by 35.746.811.1+35.7
- SimpleQA VerifiedKimi K3 by 24.650.626+24.6
- AA-Omniscience · AccuracyKimi K3 by 16.847.630.8+16.8
MathKimi K3
- FrontierMath Tiers 1-3 (v2)Kimi K3 by 4172.231.2+41
- HMMT Feb. 2026Kimi K3 by 9.19787.9+9.1
- AIME 2026Kimi K3 by 5.496.791.3+5.4
Other results9 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.
- GDPval-AA v2Kimi K3 by 720 rating points1682962+720 rating
- AA-OmniscienceKimi K3 by 50.419.7-30.8+50.4
- AA Agentic IndexKimi K3 by 4050.610.6+40
- Toolathlon VerifiedKimi K3 by 35.876.540.7+35.8
- ToolathlonKimi K3 by 34.973.238.3+34.9
- Artificial Analysis Coding IndexKimi K3 by 2876.248.2+28
- NL2RepoKimi K3 by 25.85832.2+25.8
- AA IntelligenceKimi K3 by 25.243.618.4+25.2
- HLE (with tools)Kimi K3 by 11.559.848.3+11.5
Questions people ask
Which is better, Kimi K3 or Qwen3.5 397B A17B?
Kimi K3 wins all seven areas where both have results: coding, agents, reasoning, facts, images and charts, math and long documents. Qwen3.5 397B A17B wins none, but costs 77% less.
Which is better for coding?
Kimi K3. It wins 3 of the 3 coding tests both models report; Qwen3.5 397B A17B wins none.
Which is cheaper?
Kimi K3 costs $3.00 per million input tokens and $15.00 per million output tokens; Qwen3.5 397B A17B costs $0.60 and $3.60. That makes Qwen3.5 397B A17B about 77% cheaper for the same work.
How do you compare the two?
We use the 29 benchmark tests both models have published scores on. The verdict counts the 20 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 9 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.