Gemma 4 31B vs Kimi K2.6
Wins 0 of 8 areas
—
Wins 7 of 8 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math · Long documents
Kimi K2.6 is the stronger all-rounder.Gemma 4 31B is cheaper.
Scores updated · 44 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 K2.68 of 8 tests
- ReasoningHard problems that need careful thinking03Kimi K2.63 of 3 tests
- MathCompetition and research-level math03Kimi K2.63 of 3 tests
- AgentsCarrying out multi-step tasks on its own13Kimi K2.63 of 5 tests · 1 tie
- FactsGetting facts right instead of making them up13Kimi K2.63 of 4 tests
- Images and chartsUnderstanding pictures, charts and video02Kimi K2.62 of 3 tests · 1 tie
- Long documentsFinding answers in very long texts01Kimi K2.61 of 1 test
- Following instructionsDoing exactly what it is asked00Even0 each · 1 tie
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 31B costs 89% 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 K2.6 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+44.5points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+28.2points ahead
- Fixes real GitHub issues in many programming languagesSWE-bench Multilingual+25points ahead
Where Gemma 4 31B pulls ahead
- Finds the root cause of IT system incidentsAA IT-Bench SRE+6.1points ahead
Every test, side by side
All 44 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2.6
- SWE-bench VerifiedKimi K2.6 by 28.25280.2+28.2
- SWE-bench MultilingualKimi K2.6 by 2551.776.7+25
- SWE-bench ProKimi K2.6 by 22.935.758.6+22.9
- Terminal-Bench 2.1Kimi K2.6 by 22.543.465.9+22.5
- LMArena · WebDevKimi K2.6 by 144 rating points13651509+144 rating
- LiveCodeBench v6Kimi K2.6 by 9.68089.6+9.6
- Terminal-Bench HardKimi K2.6 by 7.536.443.9+7.5
- SciCodeKimi K2.6 by 645.551.5+6
AgentsKimi K2.6
- GDPValKimi K2.6 by 20.96.127+20.9
- MCP AtlasKimi K2.6 by 10.957.268.1+10.9
- τ-Bench V3 · BankingKimi K2.6 by 8.514.823.3+8.5
- AA IT-Bench SREGemma 4 31B by 6.137.331.2+6.1
- Terminal-Bench 4.0tie00.5tie
ReasoningKimi K2.6
- Humanity's Last ExamKimi K2.6 by 13.923.637.5+13.9
- CritPtKimi K2.6 by 6.61.48+6.6
- GPQA DiamondKimi K2.6 by 5.485.791.1+5.4
FactsKimi K2.6
- AA-Omniscience · Non-hallucinationKimi K2.6 by 44.51559.5+44.5
- SimpleQA VerifiedKimi K2.6 by 24.510.434.9+24.5
- AA-Omniscience · AccuracyKimi K2.6 by 12.62032.6+12.6
- Vectara HHEM hallucination ratelower is betterGemma 4 31B by 3.47.410.8+3.4
Images and chartsKimi K2.6
- CharXiv (RQ)Kimi K2.6 by 18.867.986.7+18.8
- MMMU-ProKimi K2.6 by 673.479.4+6
- LMArena · Visiontie12771282tie
MathKimi K2.6
- HMMT Feb. 2026Kimi K2.6 by 15.577.292.7+15.5
- IMOAnswerBenchKimi K2.6 by 11.574.586+11.5
- AIME 2026Kimi K2.6 by 7.289.296.4+7.2
Other results16 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.
- Claw Eval (pass@3)Kimi K2.6 by 55.92580.9+55.9
- AA-OmniscienceKimi K2.6 by 53.2-47.95.3+53.2
- WideSearchKimi K2.6 by 45.635.280.8+45.6
- MCPMarkKimi K2.6 by 37.818.155.9+37.8
- τ²-Bench Telecom (AA run)Kimi K2.6 by 3659.995.9+36
- HLE (with tools)Kimi K2.6 by 27.526.554+27.5
- Terminal-Bench 2.0Kimi K2.6 by 23.842.966.7+23.8
- Artificial Analysis Coding IndexKimi K2.6 by 18.443.461.8+18.4
- AA Agentic IndexKimi K2.6 by 15.46.722.1+15.4
- AA IntelligenceKimi K2.6 by 12.314.727+12.3
- MATH-VisionKimi K2.6 by 7.685.693.2+7.6
- GPQA (unspecified)Kimi K2.6 by 6.784.391+6.7
- vectara_avg_summary_lengthKimi K2.6 by 40.9 rating points75.8116.7+40.9 rating
- vectara_factual_consistencyGemma 4 31B by 3.492.689.2+3.4
- MMLU-ProKimi K2.6 by 1.985.287.1+1.9
- vectara_answer_ratetie10099.7tie
Questions people ask
Which is better, Gemma 4 31B or Kimi K2.6?
Kimi K2.6 wins seven of the eight areas we test: coding, agents, reasoning, facts, images and charts, math and long documents. Gemma 4 31B wins none, but costs 89% less. They are level on following instructions.
Which is better for coding?
Kimi K2.6. It wins 8 of the 8 coding tests both models report; Gemma 4 31B wins none.
Which is cheaper?
Gemma 4 31B costs $0.17 per million input tokens and $0.40 per million output tokens; Kimi K2.6 costs $0.95 and $4.00. That makes Gemma 4 31B about 89% cheaper for the same work.
How do you compare the two?
We use the 44 benchmark tests both models have published scores on. The verdict counts the 28 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 16 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.