Gemma 4 26B A4B vs Kimi K2.5
Wins 1 of 8 areas
Following instructions
Wins 7 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents
Kimi K2.5 is the stronger all-rounder.Gemma 4 26B A4B is cheaper and better at following instructions.
Scores updated · 51 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 software08Kimi K2.58 of 8 tests
- AgentsCarrying out multi-step tasks on its own04Kimi K2.54 of 4 tests
- Images and chartsUnderstanding pictures, charts and video03Kimi K2.53 of 4 tests · 1 tie
- ReasoningHard problems that need careful thinking03Kimi K2.53 of 3 tests
- MathCompetition and research-level math03Kimi K2.53 of 3 tests
- FactsGetting facts right instead of making them up12Kimi K2.52 of 3 tests
- Long documentsFinding answers in very long texts01Kimi K2.51 of 1 test
- Following instructionsDoing exactly what it is asked10Gemma 4 26B A4B1 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.
Gemma 4 26B A4B costs 85% 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 K2.5 pulls ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+59.4points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+55.7points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+48.6points ahead
Where Gemma 4 26B A4B pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+2.2points ahead
Every test, side by side
All 51 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.5
- SWE-bench VerifiedKimi K2.5 by 59.417.476.8+59.4
- SWE-bench MultilingualKimi K2.5 by 55.717.373+55.7
- SWE-bench ProKimi K2.5 by 36.913.850.7+36.9
- Terminal-Bench HardKimi K2.5 by 21.213.634.8+21.2
- SciCodeKimi K2.5 by 94049+9
- LiveCodeBench v6Kimi K2.5 by 7.977.185+7.9
- LMArena · WebDevKimi K2.5 by 77 rating points13591436+77 rating
- Terminal-Bench 2.1Kimi K2.5 by 6.73945.7+6.7
AgentsKimi K2.5
- BrowseCompKimi K2.5 by 48.626.374.9+48.6
- GDPValKimi K2.5 by 13.83.417.2+13.8
- MCP AtlasKimi K2.5 by 13.85063.8+13.8
- τ-Bench V3 · BankingKimi K2.5 by 2.21214.2+2.2
ReasoningKimi K2.5
- Humanity's Last ExamKimi K2.5 by 11.419.330.7+11.4
- GPQA DiamondKimi K2.5 by 8.779.287.9+8.7
- CritPtKimi K2.5 by 3.103.1+3.1
FactsKimi K2.5
- AA-Omniscience · Non-hallucinationKimi K2.5 by 20.713.634.3+20.7
- AA-Omniscience · AccuracyKimi K2.5 by 16.119.135.2+16.1
- Vectara HHEM hallucination ratelower is betterGemma 4 26B A4B by 95.214.2+9
Images and chartsKimi K2.5
- MathVistaKimi K2.5 by 10.779.490.1+10.7
- CharXiv (RQ)Kimi K2.5 by 8.56977.5+8.5
- MMMU-ProKimi K2.5 by 6.269.275.4+6.2
- LMArena · Visiontie12601269tie
MathKimi K2.5
- HMMT Feb. 2026Kimi K2.5 by 8.17987.1+8.1
- AIME 2026Kimi K2.5 by 7.588.395.8+7.5
- IMOAnswerBenchKimi K2.5 by 7.574.381.8+7.5
Following instructionsGemma 4 26B A4B
- IFBenchGemma 4 26B A4B by 2.272.470.2+2.2
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.
- τ²-Bench Telecom (AA run)Kimi K2.5 by 52.343.695.9+52.3
- AA-OmniscienceKimi K2.5 by 43.5-50.8-7.3+43.5
- WideSearchKimi K2.5 by 40.738.379+40.7
- HLE (with tools)Kimi K2.5 by 3317.250.2+33
- NL2RepoKimi K2.5 by 20.411.632+20.4
- GDPval-AA v2Kimi K2.5 by 202 rating points8071009+202 rating
- SimpleVQAKimi K2.5 by 1952.271.2+19
- τ²-BenchKimi K2.5 by 17.268.285.4+17.2
- Terminal-Bench 2.0Kimi K2.5 by 16.634.250.8+16.6
- Tool-DecathlonKimi K2.5 by 15.81227.8+15.8
- MCPMarkKimi K2.5 by 15.314.229.5+15.3
- OmniDocBench 1.5Kimi K2.5 by 14.474.488.8+14.4
- AA Agentic IndexKimi K2.5 by 10.71121.7+10.7
- vectara_factual_consistencyGemma 4 26B A4B by 994.885.8+9
- vectara_answer_rateGemma 4 26B A4B by 7.699.892.2+7.6
- Artificial Analysis Coding IndexKimi K2.5 by 7.539.346.8+7.5
- τ³-BenchKimi K2.5 by 75966+7
- AA IntelligenceKimi K2.5 by 6.816.723.5+6.8
- MMLU-ProKimi K2.5 by 4.582.687.1+4.5
- vectara_avg_summary_lengthKimi K2.5 by 44.9 rating points67.1112+44.9 rating
- HMMT Feb. 2025Kimi K2.5 by 3.791.795.4+3.7
- MATH-VisionKimi K2.5 by 1.882.484.2+1.8
- HMMT Nov. 2025Kimi K2.5 by 1.787.589.2+1.7
- τ³-Bench Bankingtie1414.2tie
Questions people ask
Which is better, Gemma 4 26B A4B or Kimi K2.5?
Kimi K2.5 wins seven of the eight areas we test: coding, agents, reasoning, facts, images and charts, math and long documents. Gemma 4 26B A4B wins following instructions, and costs 85% less.
Which is better for coding?
Kimi K2.5. It wins 8 of the 8 coding tests both models report; Gemma 4 26B A4B wins none.
Which is cheaper?
Gemma 4 26B A4B costs $0.13 per million input tokens and $0.40 per million output tokens; Kimi K2.5 costs $0.60 and $3.00. That makes Gemma 4 26B A4B about 85% cheaper for the same work.
How do you compare the two?
We use the 51 benchmark tests both models have published scores on. The verdict counts the 27 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.