GPT-OSS 20B vs GPT-OSS 120B
Side-by-side comparison of API pricing, specs, benchmarks, and capabilities
| Specification |
by Groq
2 providers:
Fireworks $0.070/$0.300
Groq $0.075/$0.300
|
by Groq
3 providers:
Fireworks $0.150/$0.600
Groq $0.150/$0.600
Together $0.150/$0.600
|
|---|---|---|
| Overview | ||
| Status | Current fast Open weights | Current fast Open weights |
| Released | Jan 1, 2026 | Jan 1, 2026 |
| Pricing per million tokens | ||
| Input | $0.075/Mtok | $0.150/Mtok |
| Output | $0.300/Mtok | $0.600/Mtok |
| Blended avg | $0.131/Mtok | $0.262/Mtok |
| Specifications | ||
| Context window | 131K tokens | 131K tokens |
| Parameters | 20B | 120B |
| Speed (TPS) | 1000 tok/s | 500 tok/s |
| Modalities | ||
| Input |
text
|
text
|
| Benchmarks source: Vellum LLM Leaderboard | ||
| Avg benchmark score | 62.5 | 65.5 |
| Perf / dollar | 476.2 | 249.5 |
| GPQA Diamond | 71.5 | 80.1 |
| LiveCodeBench | 69 | 69 |
| Humanity's Last Exam | 10.9 | 14.9 |
| AIME 2025 | 98.7 | 97.9 |
| Independent composite scores each on its own scale — not part of the average above | ||
| AA Intelligence Index | 15.2 | 24.1 |
| AA Agentic Index | 3.1 | 13.4 |
| AA-Omniscience Index (−100–100) | -63 | -49.2 |
| Providers | ||
| Available from |
Fireworks — $0.070/$0.300/Mtok
Groq — $0.075/$0.300/Mtok
|
Fireworks — $0.150/$0.600/Mtok
Groq — $0.150/$0.600/Mtok
Together — $0.150/$0.600/Mtok
|
Cost at scale
1M tokens · 50/50 input/output| Volume | GPT-OSS 20B | GPT-OSS 120B | Savings |
|---|---|---|---|
| 1M tokens | $0.13 | $0.26 | $0.13 (49.5%) |
| 10M tokens | $1.31 | $2.62 | $1.31 (49.9%) |
| 100M tokens | $13.12 | $26.25 | $13.12 (50%) |
| 1000M tokens | $131.25 | $262.5 | $131.25 (50%) |
When to pick which
distilled from the pricing and spec data above- cost dominates: $0.131/Mtok blended vs $0.262 — 50% less on the same 50/50 token mix
- latency-sensitive traffic: 1000 tok/s measured throughput vs 500
- raw capability matters more than price: 65.5 vs 62.5 average benchmark score
- provider flexibility: available from 3 providers, so you can price-shop hosts or fail over
Price movement
recorded changes since we began tracking each model · last checked Aug 22, 2026| Model | Changes | Latest move | Old → new $/M |
|---|---|---|---|
| GPT-OSS 20B | 1 | price cut on Jan 27, 2026 | $0.1/$0.5 → $0.075/$0.3/Mtok |
| GPT-OSS 120B | 1 | price cut on Jan 27, 2026 | $0.15/$0.75 → $0.15/$0.6/Mtok |
The cheaper option here — GPT-OSS 20B costs $0.131/Mtok blended on Groq.
Summary
GPT-OSS 20B by Groq costs $0.075/Mtok input and $0.300/Mtok output, with a 131K-token context window. It supports text input and is available from 2 providers.
GPT-OSS 120B by Groq costs $0.150/Mtok input and $0.600/Mtok output, with a 131K-token context window. It supports text input and is available from 3 providers.
On a blended cost basis, GPT-OSS 20B is 50% cheaper than GPT-OSS 120B.
On benchmarks, GPT-OSS 120B scores higher (65.5 vs 62.5) on average. In terms of value, GPT-OSS 20B has better performance per dollar (476.2 vs 249.5).
Note: Pricing is per million tokens. Actual costs vary with usage patterns, prompt caching, and batch discounts. Always verify against official provider pricing pages.