Compare / head-to-head
Qwen3 4B Instruct 2507vs
Seed-OSS-36B-Instruct
Seed-OSS-36B-Instruct leads 7 of 7 shared benchmarks. Seed-OSS-36B-Instruct has the larger context window (512K tokens).
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
0 – 7
Seed-OSS-36B-Instruct leads
Cheaper per token
—
list price, input + output
Larger context
Seed-OSS-36B-Instruct
512K tokens
Providers
2 providers
Qwen · ByteDance Seed
Qwen3 4B Instruct 2507
Qwen · released 2025-08-06
text
- Context
- 262K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 18 · 7 core
Seed-OSS-36B-Instruct
ByteDance Seed · released 2025-08-20
text
- Context
- 512K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 18 · 12 core
Quality
Benchmark matrix
7 shared · 11 only Qwen3 4B Instruct 2507 · 11 only Seed-OSS-36B-Instruct| Benchmark | Qwen3 4B Instruct 2507 | Seed-OSS-36B-Instruct | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 7 | ||||
| GPQA Diamond | 45.8% ↗epoch run | 71.5% ↗epoch run | -25.7 pt | Seed-OSS-36B-Instruct |
| MMLU-Pro | 69.6% ↗ | 82.7% ↗ | -13.1 pt | Seed-OSS-36B-Instruct |
| AIME25 | 47.4% ↗ | 84.7% ↗ | -37.3 pt | Seed-OSS-36B-Instruct |
| Chess Puzzles (Epoch AI run) | 4% ↗epoch run | 13% ↗epoch run | -9 pt | Seed-OSS-36B-Instruct |
| IFEval | 83.4% ↗ | 85.8% ↗ | -2.4 pt | Seed-OSS-36B-Instruct |
| OTIS Mock AIME 2024-2025 (Epoch AI run) | 52.2% ↗epoch run | 67.5% ↗epoch run | -15.3 pt | Seed-OSS-36B-Instruct |
| SuperGPQA | 42.8% ↗ | 55.7% ↗ | -12.9 pt | Seed-OSS-36B-Instruct |
| Only Qwen3 4B Instruct 2507 reports · 11 | ||||
| Aider-Polyglot | 12.9% ↗ | not reported | — | — |
| Arena-Hard v2 | 43.4% ↗ | not reported | — | — |
| BFCL-v3 | 61.9% ↗ | not reported | — | — |
| GPQA | 62% ↗ | not reported | — | — |
| HMMT25 | 31% ↗ | not reported | — | — |
| LiveBench 20241125 | 63% ↗ | not reported | — | — |
| LiveCodeBench v6 (25.02-25.05) | 35.1% ↗ | not reported | — | — |
| MMLU-Redux | 84.2% ↗ | not reported | — | — |
| MultiPL-E | 76.8% ↗ | not reported | — | — |
| TAU2-Retail | 40.4% ↗ | not reported | — | — |
| ZebraLogic | 80.2% ↗ | not reported | — | — |
| Only Seed-OSS-36B-Instruct reports · 11 | ||||
| MMLU | not reported | 87.4% ↗ | — | — |
| AIME24 | not reported | 91.7% ↗ | — | — |
| BeyondAIME | not reported | 65% ↗ | — | — |
| GPQA-D | not reported | 71.4% ↗ | — | — |
| HLE | not reported | 10.1% ↗ | — | — |
| LiveCodeBench v6 (02/2025-05/2025) | not reported | 67.4% ↗ | — | — |
| MMMLU | not reported | 78.4% ↗ | — | — |
| Multi-SWE-Bench | not reported | 17% ↗ | — | — |
| RULER (128K) | not reported | 94.6% ↗ | — | — |
| SWE-Bench Verified (OpenHands) | not reported | 56% ↗ | — | — |
| TAU1-Retail | not reported | 70.4% ↗ | — | — |
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Qwen3 4B Instruct 2507 minus Seed-OSS-36B-Instruct in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing
Side by side
Official pricing pages and model cards| Spec | Qwen3 4B Instruct 2507 | Seed-OSS-36B-Instruct | Edge |
|---|---|---|---|
| Context window | 262K tokens | 512K tokens | Seed-OSS-36B-Instruct |
| Max output | — | — | — |
| Input price / 1M | — | — | — |
| Output price / 1M | — | — | — |
| Cached input / 1M | — | — | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | — | — | — |
| Modalities | text | text | Tie |
| Released | 2025-08-06 | 2025-08-20 | — |
| Cited benchmark scores | 18 | 18 | — |
Reliability
Provider status
Live from /statusMore matchups
Qwen3 4B Instruct 2507 vs …
Models sharing the most benchmarksMore matchups
Seed-OSS-36B-Instruct vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Qwen3 4B Instruct 2507 and Seed-OSS-36B-Instruct on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.