GPT-5 vs Kimi K2 Thinking
Wins 4 of 7 areas
Coding · Reasoning · Long documents · Following instructions
Wins 1 of 7 areas
Math
GPT-5 wins more areas, narrowly.Kimi K2 Thinking is cheaper and better at math.
Scores updated · 50 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.
- ReasoningHard problems that need careful thinking40GPT-54 of 4 tests
- CodingWriting and fixing software31GPT-53 of 5 tests · 1 tie
- Following instructionsDoing exactly what it is asked20GPT-52 of 2 tests
- Long documentsFinding answers in very long texts10GPT-51 of 1 test
- MathCompetition and research-level math01Kimi K2 Thinking1 of 1 test
- AgentsCarrying out multi-step tasks on its own11Even1 each
- FactsGetting facts right instead of making them up11Even1 each
Math 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.
Kimi K2 Thinking costs 72% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where GPT-5 pulls ahead
- Common-sense trick questionsSimpleBench+30.4points ahead
- Keeps track of context across a multi-turn chatMulti-Challenge+14.2points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+13.4points ahead
Where Kimi K2 Thinking pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+6.6points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+5.8points ahead
- Real work tasks from 44 professionsGDPVal+3.5points ahead
Every test, side by side
All 50 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGPT-5
- SWE-bench VerifiedGPT-5 by 11.574.963.4+11.5
- SWE-bench MultilingualKimi K2 Thinking by 5.855.361.1+5.8
- LiveCodeBench v6GPT-5 by 3.98783.1+3.9
- Terminal-Bench HardGPT-5 by 1.532.631.1+1.5
- SciCodetie42.942.4tie
AgentsEven
- BrowseCompGPT-5 by 13.454.941.5+13.4
- GDPValKimi K2 Thinking by 3.52124.5+3.5
ReasoningGPT-5
- SimpleBenchGPT-5 by 30.456.726.3+30.4
- Humanity's Last ExamGPT-5 by 4.728.523.8+4.7
- CritPtGPT-5 by 3.15.72.6+3.1
- GPQA DiamondGPT-5 by 1.685.483.8+1.6
FactsEven
- AA-Omniscience · AccuracyGPT-5 by 9.440.330.9+9.4
- AA-Omniscience · Non-hallucinationKimi K2 Thinking by 6.617.824.4+6.6
Following instructionsGPT-5
- Multi-ChallengeGPT-5 by 14.269.655.4+14.2
- IFBenchGPT-5 by 573.168.1+5
Other results33 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.
- Aider-PolyglotGPT-5 by 28.98859.1+28.9
- AA Agentic IndexKimi K2 Thinking by 21.426.547.9+21.4
- AA-OmniscienceGPT-5 by 12.7-8.7-21.4+12.7
- frontiermath_tier_4_v1GPT-5 by 12.512.50+12.5
- Arena-Hard (Creative Writing)GPT-5 by 12.192.280.1+12.1
- HLE (with tools)Kimi K2 Thinking by 9.735.244.9+9.7
- HealthBenchGPT-5 by 9.267.258+9.2
- HealthBench no toolsGPT-5 by 9.267.258+9.2
- τ²-Bench Telecom (AA run)Kimi K2 Thinking by 8.284.893+8.2
- τ²-BenchGPT-5 by 8.182.474.3+8.1
- OJ-Bench (cpp)GPT-5 by 7.556.248.7+7.5
- OJ-Bench (cpp) no toolsGPT-5 by 7.556.248.7+7.5
- HMMT Feb. 2025Kimi K2 Thinking by 588.393.3+5
- LiveCodeBenchGPT-5 by 4.383.579.2+4.3
- HMMT 2025GPT-5 by 3.993.389.4+3.9
- HMMT25 no toolsGPT-5 by 3.993.389.4+3.9
- LiveCodeBenchV6 no toolsGPT-5 by 3.98783.1+3.9
- Artificial Analysis Coding IndexGPT-5 by 337.834.8+3
- HMMT25 heavy (Tools-allowed)GPT-5 by 2.510097.5+2.5
- MMLU-ProGPT-5 by 2.587.184.6+2.5
- Longform Writing eval (Kimi K2 Thinking system card)Kimi K2 Thinking by 2.471.473.8+2.4
- xbench-DeepSearchGPT-5 by 1.877.876+1.8
- swe_bench_bashGPT-5 by 1.66563.4+1.6
- FinSearchComp-T3 (Tools-allowed)GPT-5 by 1.148.547.4+1.1
- AA IntelligenceGPT-5 by 12322+1
- MMLU-Reduxtie95.394.4tie
- BrowseComp-ZHtie6362.3tie
- Terminal-Bench 2.0tie35.235.7tie
- AIME 2025tie94.694.5tie
- AIME25 no toolstie94.694.5tie
- AIME25 heavy (Tools-allowed)tie100100tie
- Arena-Hard (Hard Prompt)tie71.971.9tie
- HMMT Nov. 2025tie89.289.2tie
Questions people ask
Which is better, GPT-5 or Kimi K2 Thinking?
GPT-5 wins four of the seven areas where both have results: coding, reasoning, long documents and following instructions. Kimi K2 Thinking wins math, and costs 72% less. They are level on agents and facts.
Which is better for coding?
GPT-5. It wins 3 of the 5 coding tests both models report; Kimi K2 Thinking wins 1, and 1 is a tie.
Which is cheaper?
GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; Kimi K2 Thinking costs $0.60 and $2.50. That makes Kimi K2 Thinking about 72% cheaper for the same work.
How do you compare the two?
We use the 50 benchmark tests both models have published scores on. The verdict counts the 17 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 33 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.