gpt-oss-20b vs Kimi K2 Instruct
Wins 2 of 6 areas
Reasoning · Following instructions
Wins 3 of 6 areas
Coding · Facts · Long documents
Kimi K2 Instruct wins more areas, narrowly.gpt-oss-20b is cheaper and better at reasoning.
Scores updated · 29 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.
- FactsGetting facts right instead of making them up02Kimi K2 Instruct2 of 2 tests
- CodingWriting and fixing software23Kimi K2 Instruct3 of 5 tests
- Long documentsFinding answers in very long texts01Kimi K2 Instruct1 of 1 test
- ReasoningHard problems that need careful thinking21gpt-oss-20b2 of 3 tests
- Following instructionsDoing exactly what it is asked10gpt-oss-20b1 of 1 test
- AgentsCarrying out multi-step tasks on its own11Even1 each
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.
gpt-oss-20b costs 92% 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 Instruct pulls ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+24.4points ahead
- Reasons across sets of long documentsAA-LCR+18.3points ahead
- Fixes real GitHub issues in Python projectsSWE-bench Verified+13.4points ahead
Where gpt-oss-20b pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+23.4points ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+14.2points ahead
- Code for real scientific research problemsSciCode+8.2points ahead
Every test, side by side
All 29 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingKimi K2 Instruct
- SWE-bench VerifiedKimi K2 Instruct by 13.452.465.8+13.4
- Terminal-Bench HardKimi K2 Instruct by 12.910.623.5+12.9
- SciCodegpt-oss-20b by 8.238.930.7+8.2
- LiveCodeBench v6gpt-oss-20b by 7.36153.7+7.3
- SWE-bench MultilingualKimi K2 Instruct by 5.441.947.3+5.4
AgentsEven
- GDPValKimi K2 Instruct by 18.2018.2+18.2
- BrowseCompgpt-oss-20b by 14.228.314.1+14.2
Reasoninggpt-oss-20b
- GPQA DiamondKimi K2 Instruct by 7.968.876.7+7.9
- Humanity's Last Examgpt-oss-20b by 4.6116.4+4.6
- CritPtgpt-oss-20b by 1.41.40+1.4
FactsKimi K2 Instruct
- AA-Omniscience · Non-hallucinationKimi K2 Instruct by 24.45.930.3+24.4
- AA-Omniscience · AccuracyKimi K2 Instruct by 9.41625.4+9.4
Following instructionsgpt-oss-20b
- IFBenchgpt-oss-20b by 23.465.141.7+23.4
Other results15 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.
- AIME 2025gpt-oss-20b by 39.789.249.5+39.7
- AA-OmniscienceKimi K2 Instruct by 36.4-63-26.6+36.4
- τ²-BenchKimi K2 Instruct by 18.147.765.8+18.1
- τ²-Bench Telecom (AA run)Kimi K2 Instruct by 13.260.273.4+13.2
- AIR-Bench 2024gpt-oss-20b by 11.98674.1+11.9
- XSTestKimi K2 Instruct by 1187.298.2+11
- AA IntelligenceKimi K2 Instruct by 6.3915.3+6.3
- Artificial Analysis Coding IndexKimi K2 Instruct by 5.220.725.9+5.2
- MMLU-ProKimi K2 Instruct by 4.776.481.1+4.7
- IFEvalKimi K2 Instruct by 3.186.789.8+3.1
- GPQA (unspecified)Kimi K2 Instruct by 2.771.574.2+2.7
- HarmBenchgpt-oss-20b by 1.398.797.4+1.3
- HealthBenchKimi K2 Instruct by 1.342.543.8+1.3
- anthropic_red_teamtie99.799.3tie
- simple_safety_teststie100100tie
Questions people ask
Which is better, gpt-oss-20b or Kimi K2 Instruct?
Kimi K2 Instruct wins three of the six areas where both have results: coding, facts and long documents. gpt-oss-20b wins reasoning and following instructions, and costs 92% less. They are level on agents.
Which is better for coding?
Kimi K2 Instruct. It wins 3 of the 5 coding tests both models report; gpt-oss-20b wins 2.
Which is cheaper?
gpt-oss-20b costs $0.07 per million input tokens and $0.18 per million output tokens; Kimi K2 Instruct costs $0.60 and $2.50. That makes gpt-oss-20b about 92% cheaper for the same work.
How do you compare the two?
We use the 29 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 15 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.