Claude Opus 4.8 vs Kimi K2.6
Wins 6 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math
Wins 1 of 8 areas
Long documents
Claude Opus 4.8 is the stronger all-rounder.Kimi K2.6 is cheaper and better at long documents.
Scores updated · 61 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 Opus 4.89 of 9 tests
- AgentsCarrying out multi-step tasks on its own61Claude Opus 4.86 of 7 tests
- MathCompetition and research-level math61Claude Opus 4.86 of 7 tests
- ReasoningHard problems that need careful thinking30Claude Opus 4.83 of 4 tests · 1 tie
- FactsGetting facts right instead of making them up30Claude Opus 4.83 of 3 tests
- Images and chartsUnderstanding pictures, charts and video20Claude Opus 4.82 of 2 tests
- Long documentsFinding answers in very long texts01Kimi K2.61 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 84% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Claude Opus 4.8 pulls ahead
- Proof problems from the US Math OlympiadUSAMO 2026+45.5points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+30.5points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+22.8points ahead
Where Kimi K2.6 pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+13.8points ahead
- Reasons across sets of long documentsAA-LCR+3.3points ahead
- Olympiad math problems with short answersIMOAnswerBench+2.5points ahead
Every test, side by side
All 61 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingClaude Opus 4.8
- Terminal-Bench 2.1Claude Opus 4.8 by 18.784.665.9+18.7
- Terminal-Bench HardClaude Opus 4.8 by 14.458.343.9+14.4
- SWE-bench ProClaude Opus 4.8 by 10.669.258.6+10.6
- SWE-bench VerifiedClaude Opus 4.8 by 8.488.680.2+8.4
- SWE-bench MultilingualClaude Opus 4.8 by 7.784.476.7+7.7
- LMArena · WebDevClaude Opus 4.8 by 47 rating points15561509+47 rating
- LiveBench · Agentic CodingClaude Opus 4.8 by 3.650.546.9+3.6
- LiveBench · CodingClaude Opus 4.8 by 3.281.878.6+3.2
- SciCodeClaude Opus 4.8 by 2.954.451.5+2.9
AgentsClaude Opus 4.8
- Terminal-Bench 4.0Claude Opus 4.8 by 21.221.70.5+21.2
- GDPValClaude Opus 4.8 by 20.847.827+20.8
- MCP AtlasClaude Opus 4.8 by 14.182.268.1+14.1
- AA ApexAgentsClaude Opus 4.8 by 10.939.428.5+10.9
- τ-Bench V3 · BankingClaude Opus 4.8 by 10.934.223.3+10.9
- OSWorld-VerifiedClaude Opus 4.8 by 10.383.473.1+10.3
- BrowseCompKimi K2.6 by 284.386.3+2
ReasoningClaude Opus 4.8
- CritPtClaude Opus 4.8 by 12.920.98+12.9
- Humanity's Last ExamClaude Opus 4.8 by 11.248.737.5+11.2
- LiveBench · ReasoningClaude Opus 4.8 by 9.889.279.4+9.8
- GPQA Diamondtie9291.1tie
FactsClaude Opus 4.8
- SimpleQA VerifiedClaude Opus 4.8 by 18.15334.9+18.1
- AA-Omniscience · AccuracyClaude Opus 4.8 by 16.248.832.6+16.2
- AA-Omniscience · Non-hallucinationClaude Opus 4.8 by 1.260.759.5+1.2
Images and chartsClaude Opus 4.8
- CharXiv (RQ)Claude Opus 4.8 by 3.289.986.7+3.2
- LMArena · VisionClaude Opus 4.8 by 13 rating points12951282+13 rating
MathClaude Opus 4.8
- USAMO 2026Claude Opus 4.8 by 45.596.751.2+45.5
- FrontierMath Tier 4Claude Opus 4.8 by 30.556.125.6+30.5
- FrontierMath Tiers 1-3 (v2)Claude Opus 4.8 by 22.88057.2+22.8
- LiveBench · MathematicsClaude Opus 4.8 by 1094.384.3+10
- AIME 2026Claude Opus 4.8 by 3.610096.4+3.6
- HMMT Feb. 2026Claude Opus 4.8 by 2.895.592.7+2.8
- IMOAnswerBenchKimi K2.6 by 2.583.586+2.5
Following instructionsEven
- IFBenchKimi K2.6 by 13.862.276+13.8
- LiveBench · Instruction FollowingClaude Opus 4.8 by 7.67264.4+7.6
Other results26 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 (Elo)Claude Opus 4.8 by 408 rating points18901482+408 rating
- GDPval-AA v2Claude Opus 4.8 by 403 rating points15931190+403 rating
- AA-OmniscienceClaude Opus 4.8 by 23.528.85.3+23.5
- AA Agentic IndexClaude Opus 4.8 by 20.542.622.1+20.5
- frontiermath_tier_4_v1Claude Opus 4.8 by 16.731.314.6+16.7
- HiL-Bench (Tools-allowed)Claude Opus 4.8 by 16.635.318.7+16.6
- AA IntelligenceClaude Opus 4.8 by 14.841.827+14.8
- Artificial Analysis Coding IndexClaude Opus 4.8 by 12.574.361.8+12.5
- AgentWorldBench - MCPKimi K2.6 by 10.354.965.2+10.3
- DeepSearchQA (F1)Claude Opus 4.8 by 10.193.183+10.1
- ToolathlonClaude Opus 4.8 by 9.959.950+9.9
- Finance Agent v2Claude Opus 4.8 by 953.944.9+9
- Terminal-Bench 2.0Claude Opus 4.8 by 7.974.666.7+7.9
- WideSearchKimi K2.6 by 7.972.980.8+7.9
- AgentWorldBench - SearchClaude Opus 4.8 by 7.635.127.5+7.6
- AgentWorldBench - TerminalClaude Opus 4.8 by 6.759.252.5+6.7
- AgentWorldBench - OSClaude Opus 4.8 by 5.866.660.8+5.8
- AgentWorldBench - SWEClaude Opus 4.8 by 5.364.158.8+5.3
- LiveBenchClaude Opus 4.8 by 577.272.2+5
- livebench_languageClaude Opus 4.8 by 4.679.775.1+4.6
- AgentWorldBench - WebClaude Opus 4.8 by 4.554.750.2+4.5
- HLE (with tools)Claude Opus 4.8 by 3.957.954+3.9
- AgentWorldBench - OverallClaude Opus 4.8 by 3.256.653.4+3.2
- AgentWorldBench - AndroidClaude Opus 4.8 by 2.661.558.9+2.6
- τ²-Bench Telecom (AA run)Kimi K2.6 by 1.594.495.9+1.5
- livebench_data_analysistie6665.1tie
Questions people ask
Which is better, Claude Opus 4.8 or Kimi K2.6?
Claude Opus 4.8 wins six of the eight areas we test: coding, agents, reasoning, facts, images and charts and math. Kimi K2.6 wins long documents, and costs 84% less. They are level on following instructions.
Which is better for coding?
Claude Opus 4.8. It wins 9 of the 9 coding tests both models report; Kimi K2.6 wins none.
Which is cheaper?
Claude Opus 4.8 costs $5.00 per million input tokens and $25.00 per million output tokens; Kimi K2.6 costs $0.95 and $4.00. That makes Kimi K2.6 about 84% cheaper for the same work.
How do you compare the two?
We use the 61 benchmark tests both models have published scores on. The verdict counts the 35 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 26 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.