Compare / gpt-oss-120b vs gpt-oss-20b

gpt-oss-120b vs gpt-oss-20b

Which is cheaper, which is more capable, and which does better at each kind of work, from 23 results on 3 sources that measured both models.

OpenAIgpt-oss-120b Released 5 Aug 2025 · $0.10 in · $0.50 out
OpenAIgpt-oss-20b Released 5 Aug 2025 · $0.030 in · $0.14 out
gpt-oss-120b wins 162 too close to callgpt-oss-20b wins 5

Of the 23 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 16 November 2026 (catalogue release 2026-11-16-49bdfa045dee).

gpt-oss-120b or gpt-oss-20b? The short answer

They score about the same, so pick gpt-oss-20b: it costs 3.5 times less per token.

  • gpt-oss-20b is cheaper: $0.058 against $0.20 per million tokens, 3.5 times less.
  • gpt-oss-120b scores higher on the Artificial Analysis Intelligence Index: 11.6 against 10.0.
  • On third-party benchmarks, on the results both models have: gpt-oss-120b does better for agents and tool use, reasoning and knowledge and writing.

Head to head

11.6
Intelligence Index
10.0
$0.20
Price per million tokens, blendedlower is better
$0.058
259
Output speed, tokens a second
277
131,072
Context window, tokens
131,072

Every fact side by side

gpt-oss-120bgpt-oss-20b
Price per million tokens$0.10 in · $0.50 out$0.030 in · $0.14 out
Blended (3 in : 1 out)$0.20$0.058
Cheapest host, blended$0.060 (Near AI, FP4)$0.022 (PrimeIntellect, FP4)
Intelligence Index11.610.0
Spring Prompt overall––
Output speed, tokens a second259277
Context window131,072 tokens131,072 tokens
Released5 Aug 20255 Aug 2025
WeightsOpenOpen
Developer based inthe USthe US

Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.

Which is better for what

Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.

Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 125 of 159 130 of 159
Professional workReal tasks from banking, consulting and law, business documents and freelance projects 49 of 51 not measured
Reasoning and knowledgeHard questions across science, maths and general knowledge 118 of 187 132 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality 124 of 169 155 of 169
Sticking to the factsSummarising without inventing things, and factual answers 129 of 139 not measured

Every shared result

Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.

Measuregpt-oss-120bgpt-oss-20b
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Overall · TextArena rating, higher is better 187 · 237 of 387 measured 1,352 1,318
Business, management and finance · TextArena rating, higher is better 183 · 213 of 380 measured 1,352 1,323
Creative writing · TextArena rating, higher is better 238 · 275 of 385 measured 1,278 1,240
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 162 · 176 of 249 measured 11.6high reasoning 10.0low reasoning
GPQA Diamond% of questions, higher is better 119 · 159 of 232 measured 78.2%high reasoning 68.8%high reasoning
Humanity's Last Exam% of questions, higher is better 118 · 153 of 250 measured 19.6%high reasoning 11.0%high reasoning
UGI Leaderboard reported by UGI Leaderboard · 3 results
Writing score · UGIscore out of 100, higher is better 116 · 182 of 230 measured 38.5medium reasoning 24.6low reasoning
Requested-length error · UGI% off the requested word count, lower is better 224 · 221 of 230 measured 107.0%medium reasoning 98.0%low reasoning
Style adherence · UGIscore from 0 to 1, higher is better 217 · 187 of 230 measured 0.29medium reasoning 0.31medium reasoning
Show all 23 results (14 more)
Measuregpt-oss-120bgpt-oss-20b
Arena (formerly LMArena) reported by Arena (formerly LMArena) · 3 results
Expert prompts · TextArena rating, higher is better 188 · 220 of 338 measured 1,357 1,314
Instruction following · TextArena rating, higher is better 204 · 257 of 387 measured 1,323 1,281
Writing, literature and language · TextArena rating, higher is better 215 · 261 of 386 measured 1,309 1,275
Artificial Analysis reported by Artificial Analysis · 11 results
Terminal-Bench 2.1% of tasks, higher is better 88 · 97 of 131 measured 26.2%high reasoning 13.9%high reasoning
Terminal-Bench 4.0% of tasks, higher is better 72 · 72 of 120 measured 0.0%high reasoning 0.0%high reasoning
τ-bench banking% of tasks, higher is better 67 · 90 of 121 measured 12.8%high reasoning 7.0%high reasoning
Artificial Analysis Coding Indexindex score, higher is better 85 · 100 of 130 measured 30.4high reasoning 20.7high reasoning
IFBench% of instructions, higher is better 58 · 69 of 195 measured 69.0%high reasoning 65.1%high reasoning
Long-context reasoning (AA-LCR)% of questions, higher is better 160 · 184 of 241 measured 52.0%high reasoning 34.7%high reasoning
Output speedtokens per second, higher is better 8 · 6 of 78 measured 259high reasoning 277low reasoning
SciCode% of problems, higher is better 101 · 90 of 121 measured 34.0%high reasoning 38.9%high reasoning
Terminal-Bench Hard% of tasks, higher is better 84 · 127 of 190 measured 23.5%high reasoning 10.6%high reasoning
Time to first answer tokenseconds, lower is better 45 · 44 of 78 measured 8.22high reasoning 7.62low reasoning
τ²-bench telecom% of tasks, higher is better 86 · 96 of 193 measured 65.8%high reasoning 60.2%high reasoning

Quick answers

Is gpt-oss-120b better than gpt-oss-20b?

They score about the same, so pick gpt-oss-20b: it costs 3.5 times less per token.

Which is cheaper, gpt-oss-120b or gpt-oss-20b?

gpt-oss-20b is cheaper: $0.058 against $0.20 per million tokens, 3.5 times less.

Which is faster, gpt-oss-120b or gpt-oss-20b?

gpt-oss-20b writes about 277 tokens a second on its usual API, against 259 for gpt-oss-120b (Artificial Analysis).

Which is better for agents and tool use, gpt-oss-120b or gpt-oss-20b?

gpt-oss-120b ranks 125 of 159 models on sale for multi-step tasks with tools: support desks, coding agents, function calls, against 130 for gpt-oss-20b.

Which is better for reasoning and knowledge, gpt-oss-120b or gpt-oss-20b?

gpt-oss-120b ranks 118 of 187 models on sale for hard questions across science, maths and general knowledge, against 132 for gpt-oss-20b.

Which is better for writing, gpt-oss-120b or gpt-oss-20b?

gpt-oss-120b ranks 124 of 169 models on sale for what people prefer in blind comparisons, and judged writing quality, against 155 for gpt-oss-20b.

Run gpt-oss-120b and gpt-oss-20b on your own prompt.

Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.