Across 13 comparable benchmarks, Haiku 4.5 has better reported results on 10 and worse on 3.
Moderate evidence13 of 13 shared results countedHow we compare
Price basis: Standard rates. The price section below shows the applicable periods.
| Fact | Haiku 4.5 | gpt-oss-20b |
|---|---|---|
| Context window | 200K tokens | 128K tokens |
| Input / 1M | $1.00 | $0.030 |
| Output / 1M | $5.00 | $0.140 |
| Blended (4:1) | $1.15 | $0.052 |
| Status | Available | Available |
| Released | October 15, 2025 | August 6, 2025 |
Blended is 4:1 input to output with 90% cache hits at listed rates — set your own mix.
Mean of 8 shared benchmarks grouped by capability on a 0–100 scale. How averages are built
Haiku 4.5 ahead on 2 of 2 counted
gpt-oss-20b ahead on 1 of 1 counted
Haiku 4.5 ahead on 2 of 2 counted
Haiku 4.5 ahead on 2 of 2 counted
Haiku 4.5 ahead on 1 of 1 counted
Split 1–1 of 2 counted
No shared comparable results for Abstract reasoning, Research & search, Long context, Cybersecurity.
Published scores, with matching modes where available. Gaps show Haiku 4.5 relative to gpt-oss-20b. Reading the results
| Benchmark | Haiku 4.5 | gpt-oss-20b | Gap |
|---|---|---|---|
| SWE-bench VerifiedRepository Engineering · Accuracy | 73.3Thinking · tools · 128K | 34.0≠Thinking · no tools | +39.3 pts#47 vs #103 of 108 |
| Terminal Bench HardAgentic Development · Accuracy | 27.3Normal · tools | 10.6≠Thinking High · tools | +16.7 pts#74 vs #115 of 157 |
| AIME2025Mathematics · Accuracy | 96.3Thinking · tools · 128K | 98.7≠Thinking · tools | −2.4 pts#18 vs #13 of 86 |
| GPQA DiamondScientific Reasoning · Accuracy | 73.3Extended · no tools | 71.5≠Thinking · no tools | +1.8 pts#126 vs #130 of 200 |
| SciCodeScientific Computing · Score | 42.2Thinking · no tools | 38.9≠Thinking High · no tools | 1.08×#72 vs #79 of 98 |
| Pinch BenchTool Orchestration · Accuracy | 82.0Thinking · tools | 66.0Thinking · tools | +16.0 pts#22 vs #35 of 38 |
| PinchBench v2Tool Orchestration · Average score (%) | 67.7Reported best | 36.3Reported best | +31.4 pts#33 vs #48 of 51 |
| GDP.pdfDocuments & Charts · Task pass rate (%) | 3.8Thinking · no tools | 2.0≠Thinking High · no tools | +1.8 pts#78 vs #84 of 96 |
| τ²-BenchService Workflows · Accuracy | 33.0Normal · tools | 47.7≠Thinking · tools | −14.7 pts#39 vs #34 of 40 |
| τ³-BankingService Workflows · Score | 9.3Thinking · tools | 7.0≠Thinking High · tools | 1.33×#95 vs #99 of 120 |
| HLEKnowledge Exams · Accuracy | 9.7Extended · no tools | 17.3≠Thinking · tools | −7.6 pts#129 vs #114 of 151 |
| MMLU-ProKnowledge Exams · Accuracy | 80.0Normal · no tools | 74.0≠Thinking · no tools | +6.0 pts#62 vs #84 of 153 |
| ECICapability Indices · ECI score (capability index, higher is better) | 142Reported best | 138Reported best | 1.03×#90 vs #106 of 168 |
Repository Engineering · Accuracy
Agentic Development · Accuracy
Mathematics · Accuracy
Scientific Reasoning · Accuracy
Scientific Computing · Score
Tool Orchestration · Accuracy
Tool Orchestration · Average score (%)
Documents & Charts · Task pass rate (%)
Service Workflows · Accuracy
Service Workflows · Score
A dash means no result is available.
| Benchmark | Haiku 4.5 | gpt-oss-20b |
|---|---|---|
| AIME 2024 | — | 96.0 |
| CritPt | — | 1.4 |
| FrontierMath | 4.1 | — |
| FrontierMath - Tier 4 | 2.1 | — |
| MMLU | — | 85.3 |
| Code Migration | 7.5 | — |
| CodeForces | — | 2516 |
| LiveCodeBench | 62.0 | — |
| SWE-Bench Pro - Public | 39.5 | — |
| SWE-Bench Pro V2 Hard | 25.5 | — |
| Terminal-Bench 1.0 | 41.8 | — |
| Vibe Code Bench v1.1 | 11.4 | — |
| Claw Bench | 89.4 | — |
| MCP-Atlas | 40.2 | — |
| OSWorld 2.1 | 15.7 | — |
| ExploitBench | 13.7 | — |
| AA Intelligence Index v5.0.1 | 7.4 | — |
| LiveBench | 61.3 | — |
| LiveBench 2.0 | 46.0 | — |
| Vals Index | 22.9 | — |
| Creative Writing | — | 665 |
| IF Bench | 54.3 | — |
| Context Arena | 36.3 | — |
| MMMU | 73.2 | — |
| MMMU-Pro | 58.6 | — |
| AA-AnalystAgent | 15.0 | — |
| AutomationBench | 0.46 | — |
| EMB (Excel Modeling) | 22.7 | — |
| Finance Agent v2 | 31.0 | — |
| Harvey Lab-AA | 61.0 | — |
| Harvey's Legal Agent Benchmark | 0.83 | — |
| Legal Research Bench | 10.6 | — |
| Public Benefits Bench v1.1 | 54.3 | — |
| SAGE | 31.8 | — |
| ARC-AGI-1 | 47.7 | — |
| ARC-AGI-2 | 4.5 | — |
| BALROG | 31.2 | — |
| Vending-Bench 2 | 459 | — |
| BrowseComp | — | 28.3 |
| GAIA | 56.4 | — |
| MedCode | 32.7 | — |
| MedScribe | 85.2 | — |
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 · 5% steps
Blended price = effective input rate at 90% cache hits × 80% + output rate × 20%
Haiku 4.5 costs2,115%morethan gpt-oss-20b per million tokens — 22.2× the price.Using standard rates at this token mix.
Haiku 4.5: Standard pricing. Anthropic official pricing (platform.claude.com/docs/en/about-claude/pricing) verified 2026-11-13
gpt-oss-20b: Standard pricing. DeepInfra official models API (openai/gpt-oss-20b) verified 2026-11-13
| Price / 1M tokens | Haiku 4.5 | gpt-oss-20b | Ratio |
|---|---|---|---|
| Input | $1.00 | $0.03 | 33.3× |
| Output | $5.00 | $0.14 | 35.7× |
| Cached input | $0.10 | — | — |
| Cache write | — | — | — |
| Blended | $1.15 | $0.052 | 22.2× |
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.
| Benchmark | Haiku 4.5 | gpt-oss-20b |
|---|---|---|
| AA Intelligence Index v5.0.1 | Not reported | |
| ARC-AGI-1 | Not reported | |
| AutomationBench | $0.07 / task Average · Pass Rate: 0.46 Reported best (effort unspecified) · zapier-private-heldout | Not reported |
| LiveBench 2.0 | $0.934 / task Average · Overall score (equal-weight mean of 7 suite scores, 0-100): 46.0 Max (task-specific tools) | Not reported |
| Public Benefits Bench v1.1 | $0.2245 / task Average · Criterion pass rate (%): 54.3 Unspecified Reasoning (With Tools) | Not reported |
| Attribute | Haiku 4.5 | gpt-oss-20b |
|---|---|---|
| Availability | Available | Available |
| Provider | Anthropic | OpenAI |
| Model type | Multimodal model | Reasoning model |
| First release | October 15, 2025 | August 6, 2025 |
| Context window | 200K tokens | 128K tokens |
| Max output | 64K tokens | 128K tokens |
| Total parameters | — | 21B |
| Active parameters | — | 3.6B |
| Architecture | Undisclosed | Mixture of Experts (MoE) |
| Documented capabilities | Reasoning, Multilingual | Reasoning, Multilingual |
| Input modalities | Text, Image | Text |
| Output modalities | Text | Text |
| Weights | Closed / API only | Open weights |
| Licence | — | Apache 2.0 |
| Commercial use | Proprietary | Commercial use allowed |
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 Haiku 4.5 or gpt-oss-20b? 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.