Compare
Four models, one page
Pick the models you are actually choosing between. Every row is scaled within itself, so a bar means something next to its neighbours and nothing across rows.
| Specification | gpt-oss-120bOpenAI | Claude Opus 5Anthropic |
|---|---|---|
| Released | Aug 5, 2025 | Jul 24, 2026 |
| Context window | 131K | 1M |
| Max output | 131K | 128K |
| Input / 1M | $0.037 | $5 |
| Output / 1M | $0.17 | $25 |
| Cost per task | $0.02 | $2.25 |
| Arena Elo | 1,034 | 1,393 |
| Serving providers | 17 | 5 |
| Parameters | 116.8B | Undisclosed |
| Licence | apache-2.0 | Proprietary |
| Capabilities |
|
|
Metrics side by side
Each row is scaled to the largest value in that row — bars compare within a row, never across rows
- gpt-oss-120b
- Claude Opus 5
Intelligence Index
gpt-oss-120b24.1Claude Opus 563.1Coding Index
gpt-oss-120b30.4Claude Opus 578.0Agentic Index
gpt-oss-120b13.4Claude Opus 559.2Output speed
gpt-oss-120b773 t/sClaude Opus 587 t/sContext window
gpt-oss-120b131KClaude Opus 51MLatency · lower is better
gpt-oss-120b205msClaude Opus 51.12sBlended price / 1M · lower is better
gpt-oss-120b$0.07Claude Opus 5$10Cost per task · lower is better
gpt-oss-120b$0.02Claude Opus 5$2.25
Rows marked “lower is better” still draw a longer bar for a larger number — read the value, not just the length. Arena Elo is in the specification table above instead: it has no meaningful zero, so a bar would flatten the gaps.
View as table
| Metric | gpt-oss-120b | Claude Opus 5 |
|---|---|---|
| Intelligence Index | 24.1 | 63.1 |
| Coding Index | 30.4 | 78.0 |
| Agentic Index | 13.4 | 59.2 |
| Output speed | 773 t/s | 87 t/s |
| Context window | 131K | 1M |
| Latency · lower is better | 205ms | 1.12s |
| Blended price / 1M · lower is better | $0.07 | $10 |
| Cost per task · lower is better | $0.02 | $2.25 |
Evaluation scores
Percentage correct on a common 0–100% scale
- gpt-oss-120b
- Claude Opus 5
GPQA Diamond
gpt-oss-120b78.2%Claude Opus 593.2%Humanity's Last Exam
gpt-oss-120b19.6%Claude Opus 554.9%SciCode
gpt-oss-120b38.9%Claude Opus 555.7%τ²-bench
gpt-oss-120b65.8%Claude Opus 542.1%Terminal-Bench Hard
gpt-oss-120b23.5%Claude Opus 5not measuredLiveCodeBench
gpt-oss-120b87.8%Claude Opus 5not measuredAA-LCR long context
gpt-oss-120b51.0%Claude Opus 575.7%AIME 2025
gpt-oss-120b93.4%Claude Opus 5not measured
A missing bar means that evaluation was not run for that model — it is not a zero.
View as table
| Evaluation | gpt-oss-120b | Claude Opus 5 |
|---|---|---|
| GPQA Diamond | 78.2% | 93.2% |
| Humanity's Last Exam | 19.6% | 54.9% |
| SciCode | 38.9% | 55.7% |
| τ²-bench | 65.8% | 42.1% |
| Terminal-Bench Hard | 23.5% | — |
| LiveCodeBench | 87.8% | — |
| AA-LCR long context | 51.0% | 75.7% |
| AIME 2025 | 93.4% | — |
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).