gpt-oss-20b vs Qwen2.5-72B-Instruct
Specs, VRAM requirements and download trends — updated 24 October 2026.
openai · A · Open weights
Params
21.5 B
Context
131,072
30d
5.7 M
apache-2.0 · commercial OK
Specification comparison
Differences are highlighted; identical values are muted.
| Specification | gpt-oss-20b | Qwen2.5-72B-Instruct |
|---|---|---|
| Deployment | Open weights · also via API | Open weights · also via API |
| Parameters | 21.5 B | 72.7 B |
| Architecture | GptOssForCausalLM | Qwen2ForCausalLM |
| Context length | 131,072 | 32,768 |
| Licence | apache-2.0 | other |
| Commercial use | Allowed | Unknown |
| Languages | — | 1 |
| Downloads 30d | 5,656,843 | 232,348 |
| Downloads all time | 105.0 M | 9.1 M |
| Quantizations on the Hub | SAFETENSORS | SAFETENSORS |
| Gated | No | No |
| First seen on the Hub | 2025-08-04 | 2024-09-16 |
| API input price / 1M tokens | $0.018 | $0.36 |
| API output price / 1M tokens | $0.09 | $0.40 |
| Est. cost per 1M input + 1M output tokens | $0.11 | $0.76 |
| Cached input / 1M tokens | $0.009 | — |
| AA Intelligence Index | 10 | 7.7 |
| AA Coding Index | — | — |
| AA Agentic Index | — | — |
| Max output tokens | 32,768 | 16,384 |
| API providers | 10 | 1 |
| Tool calling | Yes | Yes |
| Reasoning mode | Yes | No |
API cost
Estimated cost of 1M input + 1M output tokens via OpenRouter
gpt-oss-20b
$0.11
Qwen2.5-72B-Instruct
$0.76
gpt-oss-20b is 7.0× cheaper at list price.
Download trend
Daily snapshots, last 89 days (28 Jul – 24 Oct)
8.6 M232 K
gpt-oss-20b
Qwen2.5-72B-Instruct
VRAM side by side
On RTX 4090 · 24 GB · 8K context unless noted
gpt-oss-20b · fp16 @ 128K ctx
82.4 GB / 24 GB
Qwen2.5-72B-Instruct · fp16
171.4 GB / 24 GB