DeepSeek-V4-Pro vs Kimi K3
Wins 0 of 7 areas
—
Wins 7 of 7 areas
Coding · Agents · Reasoning · Facts · Math · Long documents · Following instructions
Kimi K3 is the stronger all-rounder.DeepSeek-V4-Pro is cheaper.
Scores updated · 47 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.
- AgentsCarrying out multi-step tasks on its own16Kimi K36 of 7 tests
- CodingWriting and fixing software05Kimi K35 of 5 tests
- ReasoningHard problems that need careful thinking05Kimi K35 of 5 tests
- FactsGetting facts right instead of making them up03Kimi K33 of 3 tests
- MathCompetition and research-level math13Kimi K33 of 5 tests · 1 tie
- Long documentsFinding answers in very long texts01Kimi K31 of 1 test
- Following instructionsDoing exactly what it is asked01Kimi K31 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.
DeepSeek-V4-Pro costs 93% 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 K3 pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+40.9points ahead
- Extremely hard research-level math problemsFrontierMath Tier 4+36.6points ahead
- Unpublished advanced math problemsFrontierMath Tiers 1-3 (v2)+26.9points ahead
Where DeepSeek-V4-Pro pulls ahead
- Fresh competition math problemsLiveBench · Mathematics+6.3points ahead
- Complex command-line tasks across many fieldsTerminal-Bench 4.0+2points ahead
Every test, side by side
All 47 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K3
- Terminal-Bench 2.1Kimi K3 by 216485+21
- LiveBench · Agentic CodingKimi K3 by 19.642.662.2+19.6
- LMArena · WebDevKimi K3 by 195 rating points14631658+195 rating
- LiveBench · CodingKimi K3 by 11.57081.5+11.5
- SciCodeKimi K3 by 8.750.859.5+8.7
AgentsKimi K3
- GDPValKimi K3 by 18.932.951.8+18.9
- AA ApexAgentsKimi K3 by 1724.341.3+17
- τ-Bench V3 · BankingKimi K3 by 15.930.146+15.9
- MCP AtlasKimi K3 by 10.673.684.2+10.6
- AA IT-Bench SREKimi K3 by 9.438.347.7+9.4
- BrowseCompKimi K3 by 7.883.491.2+7.8
- Terminal-Bench 4.0DeepSeek-V4-Pro by 214.612.6+2
ReasoningKimi K3
- CritPtKimi K3 by 10.512.923.4+10.5
- SimpleBenchKimi K3 by 9.850.960.7+9.8
- Humanity's Last ExamKimi K3 by 9.437.546.9+9.4
- LiveBench · ReasoningKimi K3 by 882.790.7+8
- GPQA DiamondKimi K3 by 4.788.893.5+4.7
FactsKimi K3
- AA-Omniscience · Non-hallucinationKimi K3 by 40.95.946.8+40.9
- AA-Omniscience · AccuracyKimi K3 by 4.64347.6+4.6
- SimpleQA VerifiedKimi K3 by 4.446.250.6+4.4
MathKimi K3
- FrontierMath Tier 4Kimi K3 by 36.62.439+36.6
- FrontierMath Tiers 1-3 (v2)Kimi K3 by 26.945.372.2+26.9
- LiveBench · MathematicsDeepSeek-V4-Pro by 6.390.784.4+6.3
- HMMT Feb. 2026Kimi K3 by 1.895.297+1.8
- AIME 2026tie96.796.7tie
Following instructionsKimi K3
- LiveBench · Instruction FollowingKimi K3 by 962.471.4+9
Other results20 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.
- DeepSWEKimi K3 by 54.712.867.5+54.7
- GDPval-AA v2Kimi K3 by 376 rating points13061682+376 rating
- DSBench-FullStackKimi K3 by 31.941.873.7+31.9
- DSBench-HardKimi K3 by 31.931.163+31.9
- AA-OmniscienceKimi K3 by 30.4-10.719.7+30.4
- Program BenchKimi K3 by 3047.877.8+30
- AA Agentic IndexKimi K3 by 22.927.750.6+22.9
- ToolathlonKimi K3 by 21.451.873.2+21.4
- Toolathlon VerifiedKimi K3 by 20.655.976.5+20.6
- NL2RepoKimi K3 by 19.538.558+19.5
- AutomationBench PublicKimi K3 by 1812.830.8+18
- Artificial Analysis Coding IndexKimi K3 by 16.859.476.2+16.8
- AA IntelligenceKimi K3 by 13.230.443.6+13.2
- HLE (wo / w tools)Kimi K3 by 7.848.256+7.8
- livebench_languageKimi K3 by 7.478.185.5+7.4
- Terminal-bench 3.0Kimi K3 by 5.611.817.4+5.6
- livebench_data_analysisKimi K3 by 4.274.578.7+4.2
- CyberGymDeepSeek-V4-Pro by 3.383.380+3.3
- Agents' Last ExamKimi K3 by 2.625.728.3+2.6
- HLE (with tools)tie6059.8tie
Questions people ask
Which is better, DeepSeek-V4-Pro or Kimi K3?
Kimi K3 wins all seven areas where both have results: coding, agents, reasoning, facts, math, long documents and following instructions. DeepSeek-V4-Pro wins none, but costs 93% less.
Which is better for coding?
Kimi K3. It wins 5 of the 5 coding tests both models report; DeepSeek-V4-Pro wins none.
Which is cheaper?
DeepSeek-V4-Pro costs $0.43 per million input tokens and $0.87 per million output tokens; Kimi K3 costs $3.00 and $15.00. That makes DeepSeek-V4-Pro about 93% cheaper for the same work.
How do you compare the two?
We use the 47 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 20 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.