Qwen3-30B-A3B-Instruct-2507-FP8
We introduce the updated version of the Qwen3-30B-A3B-FP8 non-thinking mode, named Qwen3-30B-A3B-Instruct-2507-FP8, featuring the following key enhancements:
Params
30.5 B
Context
262,144
Downloads 30d
498 K
Likes
132
Download history
daily snapshots · 56 days
▲ 12 K in the last 30 days (2.3%)
750 K498 K
Aug 23Sep 2Sep 12Sep 21
750 K487 K
Jul 28Aug 15Sep 3Sep 21
Can you run it?
Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.
| File | Quant | Size | Est. VRAM | Verdict on RTX 4090 · 24 GB |
|---|---|---|---|---|
| model.safetensors | f8_e4m3 | 31.2 GB | 39.4 GB | ❌ Won’t fit |
| model.safetensors (bf16, full) | bf16 + 256K ctx | 31.2 GB | 181.3 GB | ❌ Won’t fit |
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.
Run it
copy-paste, exact tags checked against the Hub$ curl -s https://aimodelscomparison.com/api/v1/models/qwen3-30b-a3b-instruct-2507-fp8
{
"hf_id": "Qwen/Qwen3-30B-A3B-Instruct-2507-FP8",
"params_b": 30.53,
"context_length": 262144,
"license": { "id": "apache-2.0", "commercial": "yes" },
"downloads_30d": 497739,
"vram_estimates": [
{ "quant": "f8_e4m3", "gb": 39.4 }
],
"updated_at": "2026-07-28T18:04:46Z"
}
Specifications
- Architecture
- Qwen3MoeForCausalLM
- Parameters
- 30.5 B
- Tensor type
- F8_E4M3
- Context length
- 262,144
- Vocabulary
- 151,936
- Layers / heads
- 48 / 32
- Licence
- apache-2.0
- First seen on the Hub
- 2025-07-28
- Base model
- Qwen3-30B-A3B-Instruct-2507
- Training datasets
- undisclosed
- Added to our catalog
- 2026-07-28
Family
Base model and the most-downloaded derivatives in the catalog.
Compare with any text-generation model