Gemini 4 Argon vs Kimi K3
Wins 6 of 7 areas
Coding · Agents · Reasoning · Facts · Images and charts · Math
Wins 1 of 7 areas
Long documents
Gemini 4 Argon is the stronger all-rounder.Kimi K3 is better at long documents.
Scores updated · 26 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 thinking40Gemini 4 Argon4 of 4 tests
- FactsGetting facts right instead of making them up30Gemini 4 Argon3 of 3 tests
- CodingWriting and fixing software20Gemini 4 Argon2 of 2 tests
- AgentsCarrying out multi-step tasks on its own20Gemini 4 Argon2 of 2 tests
- MathCompetition and research-level math20Gemini 4 Argon2 of 2 tests
- Images and chartsUnderstanding pictures, charts and video10Gemini 4 Argon1 of 1 test
- Long documentsFinding answers in very long texts01Kimi K31 of 1 test
Images and charts 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.
Gemini 4 Argon costs 33% 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 Gemini 4 Argon pulls ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+44.5points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+39.1points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+38.1points ahead
Where Kimi K3 pulls ahead
- Reasons across sets of long documentsAA-LCR+9points ahead
Every test, side by side
All 26 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingGemini 4 Argon
- SciCodeGemini 4 Argon by 2.361.859.5+2.3
- LMArena · WebDevGemini 4 Argon by 21 rating points16791658+21 rating
AgentsGemini 4 Argon
- Terminal-Bench 4.0Gemini 4 Argon by 44.557.112.6+44.5
- GDPValGemini 4 Argon by 4.556.251.7+4.5
ReasoningGemini 4 Argon
- SimpleBenchGemini 4 Argon by 3191.760.7+31
- ARC-AGI-2Gemini 4 Argon by 30.991.360.4+30.9
- Humanity's Last ExamGemini 4 Argon by 10.257.146.9+10.2
- CritPtGemini 4 Argon by 3.727.123.4+3.7
FactsGemini 4 Argon
- AA-Omniscience · Non-hallucinationGemini 4 Argon by 38.184.946.8+38.1
- SimpleQA VerifiedGemini 4 Argon by 25.47650.6+25.4
- AA-Omniscience · AccuracyGemini 4 Argon by 2.349.947.6+2.3
Images and chartsGemini 4 Argon
- MMMU-ProGemini 4 Argon by 7.888.380.5+7.8
MathGemini 4 Argon
- FrontierMath Tier 4Gemini 4 Argon by 39.17839+39.1
- FrontierMath Tiers 1-3 (v2)Gemini 4 Argon by 13.88672.2+13.8
Other results11 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 4.0Gemini 4 Argon by 44.857.412.6+44.8
- HLE-DiamondGemini 4 Argon by 29.856.726.9+29.8
- AA-OmniscienceGemini 4 Argon by 22.742.419.7+22.7
- HiL-Bench (Tools-allowed)Gemini 4 Argon by 19.36242.7+19.3
- Agents' Last ExamGemini 4 Argon by 11.239.528.3+11.2
- OSWorld 2.0Gemini 4 Argon by 10.969.258.3+10.9
- AA IntelligenceGemini 4 Argon by 952.643.6+9
- DeepSWE 1.1Gemini 4 Argon by 8.977.969+8.9
- PostTrainBenchGemini 4 Argon by 8.745.336.6+8.7
- ARC-AGI-1Gemini 4 Argon by 498.594.5+4
- GDP (Surge AI)tie21.822tie
Questions people ask
Which is better, Gemini 4 Argon or Kimi K3?
Gemini 4 Argon wins six of the seven areas where both have results: coding, agents, reasoning, facts, images and charts and math. Kimi K3 wins long documents.
Which is better for coding?
Gemini 4 Argon. It wins 2 of the 2 coding tests both models report; Kimi K3 wins none.
Which is cheaper?
Gemini 4 Argon costs $2.00 per million input tokens and $10.00 per million output tokens; Kimi K3 costs $3.00 and $15.00. That makes Gemini 4 Argon about 33% cheaper for the same work.
How do you compare the two?
We use the 26 benchmark tests both models have published scores on. The verdict counts the 15 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 11 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.