OpenAI: gpt-oss-20b
openai/gpt-oss-20b
Cheapest provider
$0.03 / 1M
DekaLLM
Fastest provider (p95)
2705 tok/s
Groq
Intelligence (estimated)
62.0
Family + generation + popularity + price tier
Per-provider performance
Latency / throughput / uptime / price measured across providers over the last 30 minutes of live traffic. This is what proves “sourced cheapest” — Atlas mode draws on these per call to serve the cheapest path that holds quality.
| Provider | Quant | Input $/1M | Output $/1M | Latency p50 / p95 | Throughput p50 / p95 | Uptime 30m | Success |
|---|---|---|---|---|---|---|---|
| DekaLLM | full· bf16 | $0.0290 | $0.1400 | 821ms / 7671ms | 71 / 123 tok/s | 90.93% | 90.6% |
| DeepInfra | full· bf16 | $0.0300 | $0.1400 | 332ms / 6487ms | 57 / 92 tok/s | 99.76% | 99.8% |
| Novita | q4· fp4 | $0.0400 | $0.1500 | 690ms / 4704ms | 133 / 166 tok/s | 99.95% | 100.0% |
| Parasail | q4· fp4 | $0.0400 | $0.2000 | 382ms / 2346ms | 131 / 218 tok/s | 99.47% | 99.4% |
| Phala | full· unknown | $0.0400 | $0.1500 | 429ms / 5210ms | 59 / 74 tok/s | 96.02% | 72.0% |
| SiliconFlow | q8· fp8 | $0.0400 | $0.1800 | 1779ms / 6708ms | 51 / 151.1 tok/s | 99.24% | 99.8% |
| Together | full· unknown | $0.0500 | $0.2000 | 376ms / 5970ms | 108 / 135.6 tok/s | 83.05% | 80.2% |
| WandB | q4· fp4 | $0.0500 | $0.2000 | 2365ms / 5340ms | 282 / 366 tok/s | 98.79% | 98.8% |
| Amazon Bedrock | full· unknown | $0.0700 | $0.1500 | 2127ms / 14830ms | 150 / 381 tok/s | — | 67.8% |
| Amazon Bedrock | full· unknown | $0.0700 | $0.1500 | 2127ms / 14830ms | 150 / 381 tok/s | — | 67.8% |
| Fireworks | full· unknown | $0.0700 | $0.3000 | 654ms / 3997ms | 80 / 161 tok/s | 98.68% | 98.6% |
| full· unknown | $0.0700 | $0.2500 | 317ms / 1256ms | 52 / 114 tok/s | 85.37% | 86.7% | |
| Groq | full· unknown | $0.0750 | $0.3000 | 247ms / 2053ms | 393.5 / 2704.7 tok/s | 99.64% | 76.4% |
| NextBit | q8· fp8 | $0.1000 | $0.4500 | 766ms / 7038ms | 152 / 216.4 tok/s | 90.49% | 88.3% |
“—” means live telemetry hasn’t accumulated enough recent traffic for that endpoint. “undisclosed” means the provider serves the model but doesn’t expose the quantization label (typically running fp8 / int8 internally).
Intelligence estimate
No public benchmark numbers indexed for this model yet, so the leaderboard score is derived from the catalogue data we sync: model family standing, generation, and live usage rank — applied identically to open- and closed-weight models. The heuristic ceiling sits below confirmed-benchmark frontier models so curated rankings stay clearly on top.
- ·openai legacy family base
Want a curated score? File a benchmark report and we’ll add it to the next sync.
How Atlas mode sources OpenAI: gpt-oss-20b
- Strict mode — pin OpenAI: gpt-oss-20b exactly and we pass it straight through, sourced from the cheapest provider above. The same model, no substitutions — currently DekaLLM at $0.03/1M.
- Atlas mode — the default. Each call is auto-optimized for the cheapest path that holds quality, at least 5% off going direct from call one and climbing as it ramps. You always see which model served the call and exactly what you saved — thumbs-down anything you don’t like for a full refund.