DeepSeek-V3.1 vs Kimi K2 (Non-Reasoning)
Wins 4 of 6 areas
Coding · Reasoning · Facts · Long documents
Wins 1 of 6 areas
Agents
DeepSeek-V3.1 is the stronger all-rounder.Kimi K2 (Non-Reasoning) is better at agents.
Scores updated · 20 tests both models report · How we compare
Where each one wins
Tests won in each of the six 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 thinking40DeepSeek-V3.14 of 4 tests
- CodingWriting and fixing software20DeepSeek-V3.12 of 2 tests
- FactsGetting facts right instead of making them up21DeepSeek-V3.12 of 3 tests
- Long documentsFinding answers in very long texts10DeepSeek-V3.11 of 1 test
- AgentsCarrying out multi-step tasks on its own01Kimi K2 (Non-Reasoning)1 of 1 test
- Following instructionsDoing exactly what it is asked00Even0 each · 1 tie
Agents, 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-V3.1 costs 22% 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 DeepSeek-V3.1 pulls ahead
- Common-sense trick questionsSimpleBench+13.7points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+9.1points ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+6.9points ahead
Where Kimi K2 (Non-Reasoning) pulls ahead
- Real work tasks from 44 professionsGDPVal+19points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+5.9points ahead
Every test, side by side
All 20 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingDeepSeek-V3.1
- Terminal-Bench HardDeepSeek-V3.1 by 9.12515.9+9.1
- SciCodeDeepSeek-V3.1 by 4.639.134.5+4.6
ReasoningDeepSeek-V3.1
- SimpleBenchDeepSeek-V3.1 by 13.74026.3+13.7
- Humanity's Last ExamDeepSeek-V3.1 by 6.914.37.4+6.9
- CritPtDeepSeek-V3.1 by 220+2
- GPQA DiamondDeepSeek-V3.1 by 1.377.976.6+1.3
FactsDeepSeek-V3.1
- Vectara HHEM hallucination ratelower is betterDeepSeek-V3.1 by 12.45.517.9+12.4
- AA-Omniscience · Non-hallucinationKimi K2 (Non-Reasoning) by 5.917.523.4+5.9
- AA-Omniscience · AccuracyDeepSeek-V3.1 by 1.62927.4+1.6
Other results8 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.
- AA Agentic IndexKimi K2 (Non-Reasoning) by 2918.947.9+29
- τ²-Bench Telecom (AA run)Kimi K2 (Non-Reasoning) by 23.737.461.1+23.7
- vectara_factual_consistencyDeepSeek-V3.1 by 12.494.582.1+12.4
- Artificial Analysis Coding IndexKimi K2 (Non-Reasoning) by 5.129.734.8+5.1
- vectara_avg_summary_lengthDeepSeek-V3.1 by 4.563.759.2+4.5
- vectara_answer_rateKimi K2 (Non-Reasoning) by 4.194.598.6+4.1
- AA-OmniscienceKimi K2 (Non-Reasoning) by 1.3-29.6-28.3+1.3
- AA Intelligencetie13.512.7tie
Questions people ask
Which is better, DeepSeek-V3.1 or Kimi K2 (Non-Reasoning)?
DeepSeek-V3.1 wins four of the six areas where both have results: coding, reasoning, facts and long documents. Kimi K2 (Non-Reasoning) wins agents. They are level on following instructions.
Which is better for coding?
DeepSeek-V3.1. It wins 2 of the 2 coding tests both models report; Kimi K2 (Non-Reasoning) wins none.
Which is cheaper?
DeepSeek-V3.1 costs $0.56 per million input tokens and $1.68 per million output tokens; Kimi K2 (Non-Reasoning) costs $0.57 and $2.30. That makes DeepSeek-V3.1 about 22% cheaper for the same work.
How do you compare the two?
We use the 20 benchmark tests both models have published scores on. The verdict counts the 12 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 8 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.