Gemini 2.5 Pro vs gpt-oss-20b
Wins 4 of 6 areas
Coding · Reasoning · Facts · Long documents
Wins 1 of 6 areas
Following instructions
Gemini 2.5 Pro is the stronger all-rounder.gpt-oss-20b is cheaper and better at following instructions.
Scores updated · 30 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.
- CodingWriting and fixing software40Gemini 2.5 Pro4 of 4 tests
- ReasoningHard problems that need careful thinking30Gemini 2.5 Pro3 of 3 tests
- FactsGetting facts right instead of making them up20Gemini 2.5 Pro2 of 2 tests
- Long documentsFinding answers in very long texts10Gemini 2.5 Pro1 of 1 test
- Following instructionsDoing exactly what it is asked01gpt-oss-20b1 of 1 test
- AgentsCarrying out multi-step tasks on its own11Even1 each · 2 ties
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 98% 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 2.5 Pro pulls ahead
- Reasons across sets of long documentsAA-LCR+34.3points ahead
- Answers hard knowledge questions correctlyAA-Omniscience · Accuracy+23.1points ahead
- Hard command-line tasks in a real terminalTerminal-Bench Hard+15.9points ahead
Where gpt-oss-20b pulls ahead
- Finds hard-to-locate facts by browsing the webBrowseComp+18.4points ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+16.4points 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.
CodingGemini 2.5 Pro
- Terminal-Bench HardGemini 2.5 Pro by 15.926.510.6+15.9
- Terminal-Bench 2.1Gemini 2.5 Pro by 14.628.513.9+14.6
- SWE-bench VerifiedGemini 2.5 Pro by 10.863.252.4+10.8
- SciCodeGemini 2.5 Pro by 7.446.338.9+7.4
AgentsEven
- BrowseCompgpt-oss-20b by 18.49.928.3+18.4
- τ-Bench V3 · BankingGemini 2.5 Pro by 2.79.77+2.7
- GDPValtie00tie
- Terminal-Bench 4.0tie00tie
ReasoningGemini 2.5 Pro
- GPQA DiamondGemini 2.5 Pro by 14.883.668.8+14.8
- Humanity's Last ExamGemini 2.5 Pro by 71811+7
- CritPtGemini 2.5 Pro by 1.22.61.4+1.2
FactsGemini 2.5 Pro
- AA-Omniscience · AccuracyGemini 2.5 Pro by 23.13916+23.1
- AA-Omniscience · Non-hallucinationGemini 2.5 Pro by 3.29.15.9+3.2
Following instructionsgpt-oss-20b
- IFBenchgpt-oss-20b by 16.448.765.1+16.4
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.
- AA-OmniscienceGemini 2.5 Pro by 46.7-16.4-63+46.7
- HarmBenchgpt-oss-20b by 33.365.498.7+33.3
- Artificial Analysis Coding IndexGemini 2.5 Pro by 2646.720.7+26
- AIR-Bench 2024gpt-oss-20b by 12.473.686+12.4
- TAU-bench (retail)Gemini 2.5 Pro by 12.26754.8+12.2
- XSTestGemini 2.5 Pro by 11.598.787.2+11.5
- MMLU-ProGemini 2.5 Pro by 9.68676.4+9.6
- τ²-BenchGemini 2.5 Pro by 6.35447.7+6.3
- AIME 2025gpt-oss-20b by 6.28389.2+6.2
- τ²-Bench Telecom (AA run)gpt-oss-20b by 6.154.160.2+6.1
- AA IntelligenceGemini 2.5 Pro by 6159+6
- HMMT Feb. 2025Gemini 2.5 Pro by 4.180.876.7+4.1
- simple_safety_testsgpt-oss-20b by 397100+3
- bbqtie96.496.7tie
- anthropic_red_teamtie99.599.7tie
Questions people ask
Which is better, Gemini 2.5 Pro or gpt-oss-20b?
Gemini 2.5 Pro wins four of the six areas where both have results: coding, reasoning, facts and long documents. gpt-oss-20b wins following instructions, and costs 98% less. They are level on agents.
Which is better for coding?
Gemini 2.5 Pro. It wins 4 of the 4 coding tests both models report; gpt-oss-20b wins none.
Which is cheaper?
Gemini 2.5 Pro costs $1.25 per million input tokens and $10.00 per million output tokens; gpt-oss-20b costs $0.07 and $0.18. That makes gpt-oss-20b about 98% 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 15 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.