GPT-5 vs Kimi K2.5
Wins 3 of 8 areas
Agents · Facts · Following instructions
Wins 4 of 8 areas
Coding · Reasoning · Images and charts · Math
Kimi K2.5 wins more areas, narrowly.GPT-5 is better at following instructions.
Scores updated · 50 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 software17Kimi K2.57 of 8 tests
- ReasoningHard problems that need careful thinking13Kimi K2.53 of 4 tests
- Images and chartsUnderstanding pictures, charts and video12Kimi K2.52 of 3 tests
- MathCompetition and research-level math01Kimi K2.51 of 1 test
- Following instructionsDoing exactly what it is asked20GPT-52 of 2 tests
- FactsGetting facts right instead of making them up21GPT-52 of 4 tests · 1 tie
- AgentsCarrying out multi-step tasks on its own21GPT-52 of 3 tests
- Long documentsFinding answers in very long texts00Even0 each · 1 tie
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 K2.5 costs 68% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Kimi K2.5 pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+20points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+17.7points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+16.5points ahead
Where GPT-5 pulls ahead
- Short factual questions, answered correctlySimpleQA Verified+15.8points ahead
- Keeps track of context across a multi-turn chatMulti-Challenge+8.2points ahead
- Customer-service tasks in a simulated bankτ-Bench V3 · Banking+7.9points 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.
CodingKimi K2.5
- SWE-bench MultilingualKimi K2.5 by 17.755.373+17.7
- Terminal-Bench 2.1Kimi K2.5 by 10.535.245.7+10.5
- SWE-bench ProKimi K2.5 by 8.941.850.7+8.9
- SciCodeKimi K2.5 by 6.142.949+6.1
- Terminal-Bench HardKimi K2.5 by 2.232.634.8+2.2
- LiveCodeBench v6GPT-5 by 28785+2
- SWE-bench VerifiedKimi K2.5 by 1.974.976.8+1.9
- LMArena · WebDevKimi K2.5 by 18 rating points14181436+18 rating
AgentsGPT-5
- BrowseCompKimi K2.5 by 2054.974.9+20
- τ-Bench V3 · BankingGPT-5 by 7.922.114.2+7.9
- GDPValGPT-5 by 3.82117.2+3.8
ReasoningKimi K2.5
- CritPtGPT-5 by 2.65.73.1+2.6
- GPQA DiamondKimi K2.5 by 2.585.487.9+2.5
- Humanity's Last ExamKimi K2.5 by 2.228.530.7+2.2
- ARC-AGI-2Kimi K2.5 by 1.99.911.8+1.9
FactsGPT-5
- AA-Omniscience · Non-hallucinationKimi K2.5 by 16.517.834.3+16.5
- SimpleQA VerifiedGPT-5 by 15.850.134.3+15.8
- AA-Omniscience · AccuracyGPT-5 by 5.140.335.2+5.1
- Vectara HHEM hallucination ratelower is bettertie15.114.2tie
Images and chartsKimi K2.5
- LMArena · VisionKimi K2.5 by 61 rating points12081269+61 rating
- CharXiv (RQ)GPT-5 by 3.681.177.5+3.6
- MMMU-ProKimi K2.5 by 1.274.275.4+1.2
Following instructionsGPT-5
- Multi-ChallengeGPT-5 by 8.269.661.4+8.2
- IFBenchGPT-5 by 2.973.170.2+2.9
Other results24 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.
- Terminal-Bench 2.0Kimi K2.5 by 15.635.250.8+15.6
- HLE (with tools)Kimi K2.5 by 1535.250.2+15
- τ²-Bench Telecom (AA run)Kimi K2.5 by 11.184.895.9+11.1
- Artificial Analysis Coding IndexKimi K2.5 by 937.846.8+9
- frontiermath_tier_4_v1GPT-5 by 8.312.54.2+8.3
- vectara_answer_rateGPT-5 by 7.799.992.2+7.7
- HMMT Feb. 2025Kimi K2.5 by 7.188.395.4+7.1
- swe_bench_bashKimi K2.5 by 5.86570.8+5.8
- vectara_avg_summary_lengthGPT-5 by 50.7 rating points162.7112+50.7 rating
- AA Agentic IndexGPT-5 by 4.826.521.7+4.8
- τ²-BenchKimi K2.5 by 382.485.4+3
- HMMT 2025Kimi K2.5 by 2.193.395.4+2.1
- VideoMMMUKimi K2.5 by 284.686.6+2
- matharena_visual_math_overallKimi K2.5 by 1.878.880.6+1.8
- AIME 2025Kimi K2.5 by 1.594.696.1+1.5
- AA-OmniscienceKimi K2.5 by 1.4-8.7-7.3+1.4
- OJ-Bench (cpp)Kimi K2.5 by 1.256.257.4+1.2
- xbench-DeepSearchGPT-5 by 1.177.876.7+1.1
- vectara_factual_consistencytie84.985.8tie
- BrowseComp-ZHtie6362.3tie
- AA Intelligencetie2323.5tie
- ARC-AGI-1tie65.765.3tie
- HMMT Nov. 2025tie89.289.2tie
- MMLU-Protie87.187.1tie
Questions people ask
Which is better, GPT-5 or Kimi K2.5?
Kimi K2.5 wins four of the eight areas we test: coding, reasoning, images and charts and math. GPT-5 wins agents, facts and following instructions. They are level on long documents.
Which is better for coding?
Kimi K2.5. It wins 7 of the 8 coding tests both models report; GPT-5 wins 1.
Which is cheaper?
GPT-5 costs $1.25 per million input tokens and $10.00 per million output tokens; Kimi K2.5 costs $0.60 and $3.00. That makes Kimi K2.5 about 68% 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 26 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 24 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.