Kimi K2.6 vs Qwen3.5 397B A17B
Wins 7 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents
Wins 1 of 8 areas
Following instructions
Kimi K2.6 is the stronger all-rounder.Qwen3.5 397B A17B is cheaper and better at following instructions.
Scores updated · 68 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.
- CodingWriting and fixing software80Kimi K2.68 of 8 tests
- MathCompetition and research-level math50Kimi K2.65 of 5 tests
- AgentsCarrying out multi-step tasks on its own41Kimi K2.64 of 6 tests · 1 tie
- ReasoningHard problems that need careful thinking30Kimi K2.63 of 3 tests
- FactsGetting facts right instead of making them up30Kimi K2.63 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20Kimi K2.62 of 2 tests
- Long documentsFinding answers in very long texts10Kimi K2.61 of 1 test
- Following instructionsDoing exactly what it is asked01Qwen3.5 397B A17B1 of 1 test
Long documents 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.
Qwen3.5 397B A17B costs 15% 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 K2.6 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+48.4points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+26points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+17.3points ahead
Where Qwen3.5 397B A17B pulls ahead
- Finds the root cause of IT system incidentsAA IT-Bench SRE+2.9points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+2.8points ahead
Every test, side by side
All 68 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.6
- Terminal-Bench 2.1Kimi K2.6 by 14.665.951.3+14.6
- LMArena · WebDevKimi K2.6 by 109 rating points15091400+109 rating
- SWE-bench ProKimi K2.6 by 7.758.650.9+7.7
- SWE-bench MultilingualKimi K2.6 by 7.476.769.3+7.4
- SciCodeKimi K2.6 by 6.751.544.8+6.7
- LiveCodeBench v6Kimi K2.6 by 689.683.6+6
- SWE-bench VerifiedKimi K2.6 by 3.880.276.4+3.8
- Terminal-Bench HardKimi K2.6 by 343.940.9+3
AgentsKimi K2.6
- BrowseCompKimi K2.6 by 17.386.369+17.3
- AA ApexAgentsKimi K2.6 by 13.228.515.3+13.2
- GDPValKimi K2.6 by 12.22714.8+12.2
- τ-Bench V3 · BankingKimi K2.6 by 9.923.313.4+9.9
- AA IT-Bench SREQwen3.5 397B A17B by 2.931.234.1+2.9
- Terminal-Bench 4.0tie0.50tie
ReasoningKimi K2.6
- Humanity's Last ExamKimi K2.6 by 8.537.529+8.5
- CritPtKimi K2.6 by 6.381.7+6.3
- GPQA DiamondKimi K2.6 by 1.891.189.3+1.8
FactsKimi K2.6
- AA-Omniscience · Non-hallucinationKimi K2.6 by 48.459.511.1+48.4
- SimpleQA VerifiedKimi K2.6 by 8.934.926+8.9
- AA-Omniscience · AccuracyKimi K2.6 by 1.832.630.8+1.8
Images and chartsKimi K2.6
- MMMU-ProKimi K2.6 by 2.179.477.3+2.1
- LMArena · VisionKimi K2.6 by 18 rating points12821264+18 rating
MathKimi K2.6
- FrontierMath Tiers 1-3 (v2)Kimi K2.6 by 2657.231.2+26
- USAMO 2026Kimi K2.6 by 14.951.236.3+14.9
- AIME 2026Kimi K2.6 by 5.196.491.3+5.1
- IMOAnswerBenchKimi K2.6 by 5.18680.9+5.1
- HMMT Feb. 2026Kimi K2.6 by 4.892.787.9+4.8
Following instructionsQwen3.5 397B A17B
- IFBenchQwen3.5 397B A17B by 2.87678.8+2.8
Other results39 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.
- AA-OmniscienceKimi K2.6 by 365.3-30.8+36
- GDPval-AA v2Kimi K2.6 by 228 rating points1190962+228 rating
- Apex-Shortlist (no tools)Kimi K2.6 by 1677.461.4+16
- IOI 2025Kimi K2.6 by 143.7 rating points585441.3+143.7 rating
- Terminal-Bench 2.0Kimi K2.6 by 14.266.752.5+14.2
- Artificial Analysis Coding IndexKimi K2.6 by 13.661.848.2+13.6
- Apex-Shortlist (with tools)Kimi K2.6 by 12.873.260.4+12.8
- ToolathlonKimi K2.6 by 11.75038.3+11.7
- AA Agentic IndexKimi K2.6 by 11.522.110.6+11.5
- MCPMarkKimi K2.6 by 9.855.946.1+9.8
- TauBench V3 - AirlineKimi K2.6 by 9.385.876.5+9.3
- IMOAnswerBench (with tools)Kimi K2.6 by 9.293.784.5+9.2
- AA IntelligenceKimi K2.6 by 8.62718.4+8.6
- SWEBench Pro PublicKimi K2.6 by 7.758.650.9+7.7
- Vals.ai Financial Agent 1.1 - without web searchQwen3.5 397B A17B by 7.35461.3+7.3
- τ³-Bench BankingKimi K2.6 by 7.220.613.4+7.2
- WideSearchKimi K2.6 by 6.880.874+6.8
- CritPt (no tools)Kimi K2.6 by 6.79.12.4+6.7
- HLE (with tools)Kimi K2.6 by 5.75448.3+5.7
- AgentWorldBench - SWEQwen3.5 397B A17B by 5.658.864.4+5.6
- TauBench V3 - RetailQwen3.5 397B A17B by 5.682.988.5+5.6
- IFBench (prompt loose) (Tools-allowed)Qwen3.5 397B A17B by 4.573.778.2+4.5
- AgentWorldBench - AndroidKimi K2.6 by 458.954.9+4
- SciCode (subtask)Kimi K2.6 by 45248+4
- GPQA (unspecified)Kimi K2.6 by 3.99187.1+3.9
- PinchBenchKimi K2.6 by 3.690.286.6+3.6
- AgentWorldBench - SearchQwen3.5 397B A17B by 3.327.530.8+3.3
- AgentWorldBench - MCPQwen3.5 397B A17B by 3.165.268.3+3.1
- ProfBench (Search)Kimi K2.6 by 35653+3
- AgentWorldBench - TerminalQwen3.5 397B A17B by 2.852.555.3+2.8
- AgentWorldBench - WebKimi K2.6 by 1.650.248.5+1.6
- MMLU-ProX (avg en/de/fr/es/it/ja/zh/hi/pt/ko) (Tools-allowed)Qwen3.5 397B A17B by 1.48586.4+1.4
- TauBench V3 - AverageKimi K2.6 by 1.472.471+1.4
- AgentWorldBench - OverallQwen3.5 397B A17B by 1.353.454.7+1.3
- MMLU-Protie87.187.8tie
- StrongREJECTtie99.899.4tie
- τ²-Bench Telecom (AA run)tie95.995.6tie
- Vals.ai Financial Agent 1.1 - with web searchtie58.859tie
- AgentWorldBench - OStie60.860.9tie
Questions people ask
Which is better, Kimi K2.6 or Qwen3.5 397B A17B?
Kimi K2.6 wins seven of the eight areas we test: coding, agents, reasoning, facts, images and charts, math and long documents. Qwen3.5 397B A17B wins following instructions, and costs 15% less.
Which is better for coding?
Kimi K2.6. It wins 8 of the 8 coding tests both models report; Qwen3.5 397B A17B wins none.
Which is cheaper?
Kimi K2.6 costs $0.95 per million input tokens and $4.00 per million output tokens; Qwen3.5 397B A17B costs $0.60 and $3.60. That makes Qwen3.5 397B A17B about 15% cheaper for the same work.
How do you compare the two?
We use the 68 benchmark tests both models have published scores on. The verdict counts the 29 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 39 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.