Model comparison
GPT-OSS 20B vs o3-mini
Head-to-head evidence from 5 shared benchmark results across 3 categories. Overall scores shown here use the public BenchAlign v5 ranking lane.
Public leaderboard positions: GPT-OSS 20B #167 (Supported); o3-mini #141 (Supported). Intervals and evidence labels describe ranking uncertainty, not a guarantee for a specific workload.
Evidence parity. GPT-OSS 20B and o3-mini share 5 comparable benchmark results. 0 of 8 categories are comparable. 13 results are unique to GPT-OSS 20B; 5 to o3-mini.
Updated July 24, 2026- Shared results
- 5
- GPT-OSS 20B only
- 13
- o3-mini only
- 5
- Comparable categories
- 0 / 8
Benchmark data for GPT-OSS 20B and o3-mini is coming soon on BenchLM.
Confidence note. This is a partial-evidence comparison with 5 shared benchmark results across 3 evidence categories; 0 of 8 categories currently have scoreable aggregates for both models. Treat the verdict as directional until coverage is more balanced.
Why this result
BenchLM has partial data for these models, but not enough overlapping benchmark coverage to produce a fair score-level comparison yet.
o3-mini is priced at $1.10 input / $4.40 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for GPT-OSS 20B. o3-mini has the larger context window at 200K, compared with 128K for GPT-OSS 20B.
Category breakdown
Exact category averages are shown below. Not measured means BenchLM does not have enough sourced public coverage for that model and category.
| Category | GPT-OSS 20B | Δ | o3-mini |
|---|---|---|---|
| Coding | GPT-OSS 20BNot measured | MarginNo overlap | o3-mini49.3 |
| Knowledge | GPT-OSS 20BNot measured | MarginNo overlap | o3-mini77.2 |
| Inst. Following | GPT-OSS 20BNot measured | MarginNo overlap | o3-mini93.9 |
Operational comparison
Runtime and commercial metrics are compared only when both models have a complete sourced value.
| Metric | GPT-OSS 20B | o3-mini | Comparison |
|---|---|---|---|
| Input / output priceUSD per 1M tokens | GPT-OSS 20B$0 input / $0 output | o3-mini$1.1 input / $4.4 output | GPT-OSS 20B has the lower combined listed price. |
| Generation speedtokens per second | GPT-OSS 20B313 tok/s | o3-mini160 tok/s | GPT-OSS 20B has the higher measured throughput. |
| First-answer latencyseconds to first token | GPT-OSS 20B0.65 s | o3-mini7.12 s | GPT-OSS 20B reaches the first token sooner. |
| Context windowmaximum listed tokens | GPT-OSS 20B128K | o3-mini200K | o3-mini lists the larger context window. |
Benchmark Deep Dive
Agentic5 benchmarks
Coding4 benchmarks
Reasoning2 benchmarks
Knowledge8 benchmarks
| Benchmark | GPT-OSS 20B | o3-mini | Result |
|---|---|---|---|
| Artificial Analysis Intelligence IndexSource | 14.9% | 19.0% | o3-mini leads |
| AA-GPQA DiamondSource | 68.8% | 74.8% | o3-mini leads |
| AA-HLESource | 9.8% | 8.7% | GPT-OSS 20B leads |
| AA-Omniscience IndexSource | -63.9% | — | Not comparable |
| AA-Omniscience AccuracySource | 15.5% | — | Not comparable |
| AA-Omniscience Hallucination RateSource | 94.1% | — | Not comparable |
| MMLUSource | — | 86.9% | Not comparable |
| GPQASource | — | 77.2% | Not comparable |
Math1 benchmarks
| Benchmark | GPT-OSS 20B | o3-mini | Result |
|---|---|---|---|
| AIME 2024Source | — | 87.3% | Not comparable |
Multimodal1 benchmarks
| Benchmark | GPT-OSS 20B | o3-mini | Result |
|---|---|---|---|
| Design Arena WebsiteSource | 882 | — | Not comparable |
Frequently Asked Questions (3)
Can I compare GPT-OSS 20B and o3-mini on BenchLM yet?
Not fully yet. BenchLM is tracking both models, but the sourced benchmark breakdown for this comparison is still coming soon.
Why does this comparison show “coming soon”?
BenchLM only shows category winners and benchmark-level calls when we have sourced results that can be compared fairly. For these models, the public benchmark coverage is not complete enough yet.
What data is available for GPT-OSS 20B and o3-mini today?
GPT-OSS 20B: $0.00 input / $0.00 output per 1M tokens o3-mini: $1.10 input / $4.40 output per 1M tokens Both model pages still include creator, context window, reasoning mode, and other metadata while benchmark coverage fills in.