Kimi K2 Thinking vs Step 3.5 Flash
Wins 4 of 7 areas
Reasoning · Facts · Long documents · Following instructions
Wins 2 of 7 areas
Agents · Math
Kimi K2 Thinking wins more areas, narrowly.Step 3.5 Flash is cheaper and better at agents.
Scores updated · 30 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.
- FactsGetting facts right instead of making them up20Kimi K2 Thinking2 of 2 tests
- ReasoningHard problems that need careful thinking10Kimi K2 Thinking1 of 3 tests · 2 ties
- Long documentsFinding answers in very long texts10Kimi K2 Thinking1 of 1 test
- Following instructionsDoing exactly what it is asked10Kimi K2 Thinking1 of 1 test
- AgentsCarrying out multi-step tasks on its own02Step 3.5 Flash2 of 2 tests
- MathCompetition and research-level math01Step 3.5 Flash1 of 1 test
- CodingWriting and fixing software22Even2 each
Math, 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.
Step 3.5 Flash costs 87% less for the same work.
The biggest differences
The three tests each model wins by the widest margin. Scores are out of 100.
Where Kimi K2 Thinking pulls ahead
- Reasons across sets of long documentsAA-LCR+22points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+10.1points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+7.3points ahead
Where Step 3.5 Flash pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+27.5points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+11points ahead
- Olympiad math problems with short answersIMOAnswerBench+6.8points ahead
Every test, side by side
All 30 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- SWE-bench VerifiedStep 3.5 Flash by 1163.474.4+11
- Terminal-Bench HardKimi K2 Thinking by 3.831.127.3+3.8
- LiveCodeBench v6Step 3.5 Flash by 3.383.186.4+3.3
- SciCodeKimi K2 Thinking by 242.440.4+2
AgentsStep 3.5 Flash
- BrowseCompStep 3.5 Flash by 27.541.569+27.5
- GDPValStep 3.5 Flash by 3.124.527.6+3.1
ReasoningKimi K2 Thinking
- Humanity's Last ExamKimi K2 Thinking by 2.723.821.1+2.7
- GPQA Diamondtie83.883.1tie
- CritPttie2.62.5tie
FactsKimi K2 Thinking
- AA-Omniscience · Non-hallucinationKimi K2 Thinking by 10.124.414.3+10.1
- AA-Omniscience · AccuracyKimi K2 Thinking by 7.330.923.6+7.3
Following instructionsKimi K2 Thinking
- IFBenchKimi K2 Thinking by 3.568.164.6+3.5
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.
- AA-OmniscienceKimi K2 Thinking by 20.4-21.4-41.8+20.4
- xbench-DeepSearchKimi K2 Thinking by 19.77656.3+19.7
- Terminal-Bench 2.0Step 3.5 Flash by 15.335.751+15.3
- τ²-BenchStep 3.5 Flash by 13.974.388.2+13.9
- ResearchRubricsStep 3.5 Flash by 9.156.265.3+9.1
- GAIA (no file)Step 3.5 Flash by 8.975.684.5+8.9
- browsecomp_with_context_managerStep 3.5 Flash by 8.860.269+8.8
- AA IntelligenceKimi K2 Thinking by 5.42216.6+5.4
- HMMT Feb. 2025Step 3.5 Flash by 5.193.398.4+5.1
- HMMT Nov. 2025Step 3.5 Flash by 4.889.294+4.8
- BrowseComp-ZHStep 3.5 Flash by 4.662.366.9+4.6
- AA Agentic IndexStep 3.5 Flash by 4.147.952+4.1
- Artificial Analysis Coding IndexKimi K2 Thinking by 3.234.831.6+3.2
- AIME 2025Step 3.5 Flash by 2.894.597.3+2.8
- τ²-Bench Telecom (AA run)Step 3.5 Flash by 1.49394.4+1.4
- MMLU-Protie84.684.4tie
Questions people ask
Which is better, Kimi K2 Thinking or Step 3.5 Flash?
Kimi K2 Thinking wins four of the seven areas where both have results: reasoning, facts, long documents and following instructions. Step 3.5 Flash wins agents and math, and costs 87% less. They are level on coding.
Which is better for coding?
Neither. They win 2 coding tests each of the 4 both models report.
Which is cheaper?
Kimi K2 Thinking costs $0.60 per million input tokens and $2.50 per million output tokens; Step 3.5 Flash costs $0.10 and $0.30. That makes Step 3.5 Flash about 87% cheaper for the same work.
How do you compare the two?
We use the 30 benchmark tests both models have published scores on. The verdict counts the 14 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.