Across 19 comparable benchmarks, Gemini 3.1 Pro Preview has better reported results on 19 and worse on 0.
Moderate evidence19 of 19 shared results countedHow we compare
Price basis: Standard rates vs listed rates. The price section below shows the applicable periods.
| Fact | GPT OSS 120B | Gemini 3.1 Pro Preview |
|---|---|---|
| Context window | 128K tokens | 1M tokens |
| Input / 1M | $0.350 | $2.00 |
| Output / 1M | $0.750 | $12.00 |
| Blended (4:1) | $0.430 | $2.70 |
| Status | Available | Preview |
| Released | August 6, 2025 | February 20, 2026 |
Blended is 4:1 input to output with 90% cache hits at listed rates β set your own mix.
Mean of 9 shared benchmarks grouped by capability on a 0β100 scale. How averages are built
Gemini 3.1 Pro Preview ahead on 4 of 4 counted
Gemini 3.1 Pro Preview ahead on 1 of 1 counted
No fixed-scale average β see the rows in Benchmark results
Gemini 3.1 Pro Preview ahead on 3 of 3 counted
Gemini 3.1 Pro Preview ahead on 3 of 3 counted
Gemini 3.1 Pro Preview ahead on 1 of 1 counted
Gemini 3.1 Pro Preview ahead on 4 of 4 counted
No shared comparable results for Mathematics, Research & search, Long context, Cybersecurity.
Published scores, with matching modes where available. Gaps show GPT OSS 120B relative to Gemini 3.1 Pro Preview. Reading the results
| Benchmark | GPT OSS 120B | Gemini 3.1 Pro Preview | Gap |
|---|---|---|---|
| APEX-SWERepository Engineering Β· Pass@1 | 3.3Thinking High Β· tools | 33.9Thinking High Β· tools | β30.6 pts#61 vs #45 of 61 |
| LiveCodeBenchAlgorithmic Coding Β· Pass @K | 87.8Thinking High Β· no tools | 91.7β Thinking High Β· tools | 0.96Γ#14 vs #4 of 113 |
| SWE-bench VerifiedRepository Engineering Β· Accuracy | 60.1Thinking Β· no tools | 80.6β Thinking High Β· tools | β20.5 pts#78 vs #12 of 108 |
| Terminal Bench HardAgentic Development Β· Accuracy | 23.5Thinking High Β· tools | 53.8β Thinking Β· tools | β30.3 pts#88 vs #8 of 157 |
| SimpleBenchCommonsense Β· Score (AVG@5) | 22.1Thinking Β· no tools | 79.6β Normal Β· no tools | 0.28Γ#99 vs #15 of 106 |
| CritPtScientific Reasoning Β· Score | 1.1Thinking High Β· no tools | 17.7β Thinking Β· no tools | 0.06Γ#103 vs #34 of 134 |
| GPQA DiamondScientific Reasoning Β· Accuracy | 80.1Thinking Β· no tools | 94.4β Reported best | β14.3 pts#111 vs #12 of 200 |
| SciCodeScientific Computing Β· Score | 34.0Thinking High Β· no tools | 58.7β Thinking Β· no tools | 0.58Γ#88 vs #18 of 98 |
| APEX-AgentsTool Orchestration Β· Score | 4.4Reported best | 35.1Reported best | 0.13Γ#61 vs #52 of 61 |
| Pinch BenchTool Orchestration Β· Accuracy | 60.6Thinking Β· tools | 86.7Thinking Β· tools | β26.1 pts#36 vs #11 of 38 |
| PinchBench v2Tool Orchestration Β· Average score (%) | 44.8Reported best | 81.0Reported best | β36.2 pts#46 vs #13 of 51 |
| GDP.pdfDocuments & Charts Β· Task pass rate (%) | 4.0Thinking High Β· no tools | 17.8β Thinking Β· no tools | β13.8 pts#77 vs #33 of 96 |
| Creative WritingWriting | 956Normal Β· no tools | 1491Normal Β· no tools | 0.64Γ#104 vs #61 of 121 |
| AutomationBenchOffice & Business Β· Pass Rate | 0.15Thinking High Β· toolsΒ· zapier-private-heldout | 8.7Thinking High Β· toolsΒ· zapier-private-heldout | β8.5 pts#64 vs #51 of 64 |
| Harvey Lab-AALegal Β· Criterion pass rate (some entries report all-pass rate) | 13.9Thinking High Β· tools | 58.9β Thinking Β· tools | β45.0 pts#39 vs #33 of 39 |
| ΟΒ³-BankingService Workflows Β· Score | 12.8Thinking High Β· tools | 26.0Thinking High Β· tools | 0.49Γ#83 vs #46 of 120 |
| HLEKnowledge Exams Β· Accuracy | 19.0Thinking Β· tools | 51.4β Thinking High Β· tools | β32.4 pts#107 vs #44 of 151 |
| ECICapability Indices Β· ECI score (capability index, higher is better) | 140Reported best | 155Reported best | 0.90Γ#97 vs #31 of 168 |
| LiveBenchCross-capability Suites Β· Accuracy | 46.1Normal Β· no tools | 77.0β Thinking High Β· no tools | β30.9 pts#100 vs #24 of 106 |
Repository Engineering Β· Pass@1
Algorithmic Coding Β· Pass @K
Repository Engineering Β· Accuracy
Agentic Development Β· Accuracy
Commonsense Β· Score (AVG@5)
Scientific Reasoning Β· Score
Scientific Reasoning Β· Accuracy
Scientific Computing Β· Score
Tool Orchestration Β· Score
Tool Orchestration Β· Accuracy
Tool Orchestration Β· Average score (%)
Documents & Charts Β· Task pass rate (%)
Writing
Office & Business Β· Pass Rate
Legal Β· Criterion pass rate (some entries report all-pass rate)
Service Workflows Β· Score
A dash means no result is available.
Time-of-day providers use peak-hour rates; fixed-price providers use standard rates. Off-peak discounts remain in the published details.
Drag the handle above the bar for cache hits, the one below it for the input : output mix Β· 5% steps
Blended price = effective input rate at 90% cache hits Γ 80% + output rate Γ 20%
GPT OSS 120B costs84%lessthan Gemini 3.1 Pro Preview per million tokens β 6.3Γ cheaper.Using standard rates vs listed rates at this token mix.
GPT OSS 120B: Standard pricing. Cerebras Shared Inference official pricing (gpt-oss-120b) verified 2026-11-13
Gemini 3.1 Pro Preview: Standard pricing. context_length <= 200K Google Gemini API official pricing (ai.google.dev/gemini-api/docs/pricing) verified 2026-10-21; paid tier; standard tier; prompts <= 200k tokens context_length > 200K Google Gemini API official pricing (ai.google.dev/gemini-api/docs/pricing) verified 2026-10-21; paid tier; standard tier; prompts > 200k tokens
| Price / 1M tokens | GPT OSS 120B | Gemini 3.1 Pro Preview | Ratio |
|---|---|---|---|
| Input | $0.35 | $2.00 | 0.17Γ |
| Output | $0.75 | $12.00 | 0.06Γ |
| Cached input | β | $0.20 | β |
| Cache write | β | β | β |
| Blended | $0.43 | $2.70 | 0.16Γ |
Estimates use the selected token mix. Cache writes, tool fees and long-context surcharges are excluded.
Reported cost per task, paired with the score from the same run. Settings and sources may differ; costs do not affect the verdict.
GPT OSS 120B relative to Gemini 3.1 Pro Preview
Summary: 1 of 1 shared benchmarks have matching recorded settings and sources. The average is a geometric mean of 1 positive-cost ratios, one equal weight per benchmark. This benchmark set can include related tests and indexes; it is not a workload cost estimate.
GPT OSS 120B 30Γ cheaper
GPT OSS 120B 30Γ cheaper
Effort, protocol, aggregation and source match; other run conditions may differ.
GPT OSS 120B
Gemini 3.1 Pro Preview
| Benchmark | GPT OSS 120B | Gemini 3.1 Pro Preview |
|---|---|---|
| AA Intelligence Index v5.0.1 | Not reported | |
| Agents' Last Exam v1.0 | Not reported | $14.2116 / task Average Β· Pass rate: 16.4 Unspecified Reasoning (With Tools) Β· pass-rate |
| AIME 2026 | Not reported | |
| DeepSWE | Not reported | |
| IOI (Vals v2) | Not reported | |
| LiveBench 2.0 | Not reported | $0.6982 / task Average Β· Overall score (equal-weight mean of 7 suite scores, 0-100): 54.9 Thinking High (With Tools) |
| MysteryMechanism | Not reported | |
| Tax Agent Bench | Not reported | $0.3016 / task Average Β· Accuracy (%, rubric Γ citation quality): 9.3 Thinking High (With Tools) |
| Vals Index v2.1 | Not reported | |
| Vibe Code Bench 1-100 | Not reported |
| Attribute | GPT OSS 120B | Gemini 3.1 Pro Preview |
|---|---|---|
| Availability | Available | Preview |
| Provider | OpenAI | Google Deep Mind |
| Model type | Reasoning model | Multimodal model |
| First release | August 6, 2025 | February 20, 2026 |
| Context window | 128K tokens | 1M tokens |
| Max output | 128K tokens | 64K tokens |
| Total parameters | 117B | β |
| Active parameters | 5.1B | β |
| Architecture | Mixture of Experts (MoE) | Undisclosed |
| Documented capabilities | Reasoning, Multilingual | Reasoning, Multilingual |
| Knowledge cutoff | β | 2025-01 |
| Input modalities | Text | Text, Image, Audio, Video |
| Output modalities | Text | Text |
| Weights | Open weights | Closed / API only |
| Licence | Apache 2.0 | β |
| Commercial use | Commercial use allowed | Proprietary |
These are reported-score comparisons, not controlled head-to-head tests.
What people found comparing these models, or using either one. Comments also appear on the models' own pages.
Used GPT OSS 120B or Gemini 3.1 Pro Preview? Sign in with Google to share what worked, what did not, and which API or setup you tested.
No community comments on either model yet.
Alternatives with the most benchmark overlap for this matchup.