Claude Fable 5 vs Kimi K2.6
Wins 7 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents
Wins 0 of 8 areas
—
Claude Fable 5 is the stronger all-rounder.Kimi K2.6 is cheaper.
Scores updated · 53 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 software90Claude Fable 59 of 9 tests
- AgentsCarrying out multi-step tasks on its own70Claude Fable 57 of 7 tests
- ReasoningHard problems that need careful thinking40Claude Fable 54 of 4 tests
- MathCompetition and research-level math30Claude Fable 53 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20Claude Fable 52 of 2 tests
- FactsGetting facts right instead of making them up21Claude Fable 52 of 3 tests
- Long documentsFinding answers in very long texts10Claude Fable 51 of 1 test
- Following instructionsDoing exactly what it is asked11Even1 each
Long documents rests on a single test.
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.
Kimi K2.6 costs 92% 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 Claude Fable 5 pulls ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+64.6points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+41.9points ahead
- Short factual questions, answered correctlySimpleQA Verified+35.8points ahead
Where Kimi K2.6 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+23.1points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+12.5points ahead
Every test, side by side
All 53 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Fable 5
- SWE-bench ProClaude Fable 5 by 21.48058.6+21.4
- Terminal-Bench HardClaude Fable 5 by 1962.943.9+19
- Terminal-Bench 2.1Claude Fable 5 by 18.784.665.9+18.7
- LiveBench · Agentic CodingClaude Fable 5 by 15.362.246.9+15.3
- SWE-bench VerifiedClaude Fable 5 by 14.89580.2+14.8
- LMArena · WebDevClaude Fable 5 by 116 rating points16251509+116 rating
- SWE-bench MultilingualClaude Fable 5 by 9.986.676.7+9.9
- SciCodeClaude Fable 5 by 9.56151.5+9.5
- LiveBench · CodingClaude Fable 5 by 7.48678.6+7.4
AgentsClaude Fable 5
- Terminal-Bench 4.0Claude Fable 5 by 41.942.40.5+41.9
- GDPValClaude Fable 5 by 28.655.627+28.6
- MCP AtlasClaude Fable 5 by 16.684.768.1+16.6
- AA ApexAgentsClaude Fable 5 by 14.843.328.5+14.8
- τ-Bench V3 · BankingClaude Fable 5 by 14.838.123.3+14.8
- OSWorld-VerifiedClaude Fable 5 by 11.98573.1+11.9
- BrowseCompClaude Fable 5 by 1.187.486.3+1.1
ReasoningClaude Fable 5
- CritPtClaude Fable 5 by 20.628.68+20.6
- Humanity's Last ExamClaude Fable 5 by 1855.537.5+18
- LiveBench · ReasoningClaude Fable 5 by 10.389.779.4+10.3
- GPQA DiamondClaude Fable 5 by 1.592.691.1+1.5
FactsClaude Fable 5
- SimpleQA VerifiedClaude Fable 5 by 35.870.734.9+35.8
- AA-Omniscience · AccuracyClaude Fable 5 by 32.865.332.6+32.8
- AA-Omniscience · Non-hallucinationKimi K2.6 by 23.136.459.5+23.1
Images and chartsClaude Fable 5
- MMMU-ProClaude Fable 5 by 4.884.279.4+4.8
- LMArena · VisionClaude Fable 5 by 43 rating points13251282+43 rating
MathClaude Fable 5
- FrontierMath Tier 4Claude Fable 5 by 64.690.225.6+64.6
- FrontierMath Tiers 1-3 (v2)Claude Fable 5 by 29.88757.2+29.8
- LiveBench · MathematicsClaude Fable 5 by 11.79684.3+11.7
Following instructionsEven
- IFBenchKimi K2.6 by 12.563.576+12.5
- LiveBench · Instruction FollowingClaude Fable 5 by 11.475.864.4+11.4
Other results22 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 v2Claude Fable 5 by 533 rating points17231190+533 rating
- GDPval-AA (Elo)Claude Fable 5 by 450 rating points19321482+450 rating
- AA-OmniscienceClaude Fable 5 by 3843.35.3+38
- HiL-Bench (Tools-allowed)Claude Fable 5 by 37.656.318.7+37.6
- AA Agentic IndexClaude Fable 5 by 28.95122.1+28.9
- AA IntelligenceClaude Fable 5 by 22.649.627+22.6
- SWEBench Pro PublicClaude Fable 5 by 21.48058.6+21.4
- livebench_languageClaude Fable 5 by 15.690.775.1+15.6
- livebench_data_analysisClaude Fable 5 by 15.480.565.1+15.4
- Artificial Analysis Coding IndexClaude Fable 5 by 14.776.561.8+14.7
- ToolathlonClaude Fable 5 by 11.761.750+11.7
- Finance Agent v2Claude Fable 5 by 11.456.344.9+11.4
- DeepSearchQA (F1)Claude Fable 5 by 11.294.283+11.2
- HLE (with tools)Claude Fable 5 by 9.863.854+9.8
- τ³-Bench BankingClaude Fable 5 by 6.226.820.6+6.2
- LiveBenchClaude Fable 5 by 6.178.372.2+6.1
- Global-MMLU-LiteClaude Fable 5 by 4.993.388.4+4.9
- BrowseComp (w/ Ctx)Claude Fable 5 by 4.88883.2+4.8
- charxiv_rq_with_pythonClaude Fable 5 by 2.789.486.7+2.7
- τ²-Bench Telecom (AA run)Claude Fable 5 by 2.698.595.9+2.6
- StrongREJECTKimi K2.6 by 1.198.799.8+1.1
- WideSearchtie81.280.8tie
Questions people ask
Which is better, Claude Fable 5 or Kimi K2.6?
Claude Fable 5 wins seven of the eight areas we test: coding, agents, reasoning, facts, images and charts, math and long documents. Kimi K2.6 wins none, but costs 92% less. They are level on following instructions.
Which is better for coding?
Claude Fable 5. It wins 9 of the 9 coding tests both models report; Kimi K2.6 wins none.
Which is cheaper?
Claude Fable 5 costs $10.00 per million input tokens and $50.00 per million output tokens; Kimi K2.6 costs $0.95 and $4.00. That makes Kimi K2.6 about 92% cheaper for the same work.
How do you compare the two?
We use the 53 benchmark tests both models have published scores on. The verdict counts the 31 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 22 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.