Models / Clef-flash

Clef-flash

consumer
HuggingFace → HuggingBay → Cloudflare → Find for my hardware

The laptop end of the decision-model wave. Clef-flash is Cloudflare’s 9B sibling: same Qwen3.5-family backbone, same typed-answer contract (pick from a fixed set of answers, attach a confidence number, never write an essay), about 5.4GB in 4-bit so it runs on a 16GB laptop with room for the OS. 17,600 downloads in a week on Hugging Face, Apache 2.0.

Where it beats the 27B. Latency-sensitive routing: the flash build answers in one forward pass at laptop scale, which makes it the natural Worker-side classifier for abuse detection, ticket routing, and gating agent tool-calls at the edge. Anything needing the bigger model’s judgment on long documents falls back to Clef at 27B.

decision-model classification routing
Parameters
9.4B
Context
262k
License
apache 2.0
Developer
Cloudflare
Origin
🇺🇸 USA
Released
Oct 2026

Related models

Clef-Omni
35.3B 3.0B active enthusiast 24GB min RAM Oct 9, 2026
Clef
27.4B enthusiast 16GB min RAM Oct 1, 2026

Save your hardware and every model page answers the real question: will it run on your machine, and how fast?

Join free - save your rig →

Run it locally

Per-quant memory needs and a static "can you run it?" reference - no rig entry required

Q4_K_M
5.4GB 8.0GB min 16.0GB rec
Balanced - the usual local sweet spot

The reference hardware

Schematic of the Single GTX 1080 Ti (11GB) reference rig - 11GB VRAM, 484 GB/s aggregate bandwidth
Schematic of the 4x H100 80GB (320GB) reference rig - 320GB VRAM, 13400 GB/s aggregate bandwidth
Schematic of the NVIDIA DGX Station 748GB reference rig - 748GB unified memory, 8000 GB/s aggregate bandwidth
Schematic of the 8x RTX 3090 rack (192GB) reference rig - 192GB VRAM, 7489 GB/s aggregate bandwidth
Schematic of the 4x RTX 5090 (128GB) reference rig - 128GB VRAM, 7168 GB/s aggregate bandwidth
Schematic of the AMD Instinct MI300X (192GB) reference rig - 192GB VRAM, 5324 GB/s aggregate bandwidth
Schematic of the 4x RTX 4090 (96GB) reference rig - 96GB VRAM, 4032 GB/s aggregate bandwidth
Schematic of the 2x RTX 5090 (64GB) reference rig - 64GB VRAM, 3584 GB/s aggregate bandwidth
Schematic of the 2x RTX 3090 (48GB) reference rig - 48GB VRAM, 1872 GB/s aggregate bandwidth
Schematic of the Single RTX 5090 (32GB) reference rig - 32GB VRAM, 1792 GB/s aggregate bandwidth
Schematic of the RTX PRO 6000 Blackwell (96GB) reference rig - 96GB VRAM, 1792 GB/s aggregate bandwidth
Schematic of the Mac Studio M4 Ultra 192GB reference rig - 192GB unified memory, 1092 GB/s aggregate bandwidth
Schematic of the Mac Studio M4 Ultra 512GB reference rig - 512GB unified memory, 1092 GB/s aggregate bandwidth
Schematic of the Single RTX 4090 (24GB) reference rig - 24GB VRAM, 1008 GB/s aggregate bandwidth
Schematic of the MacBook Pro M5 Max 128GB reference rig - 128GB unified memory, 614 GB/s aggregate bandwidth
Schematic of the Dual EPYC 9004 + 768GB DDR5-4800 reference rig - 768GB unified memory, 460 GB/s aggregate bandwidth
Schematic of the DGX Spark 128GB unified reference rig - 128GB unified memory, 273 GB/s aggregate bandwidth
Schematic of the Ryzen AI Max+ 395 128GB reference rig - 128GB unified memory, 256 GB/s aggregate bandwidth
Schematic of the Jetson AGX Orin 64GB reference rig - 64GB unified memory, 204 GB/s aggregate bandwidth
Schematic of the Epyc + 512GB DDR4-3200 + 2x RTX 3090 reference rig - 560GB unified memory, 204 GB/s aggregate bandwidth
Schematic of the Epyc + 512GB DDR4-2400 + 2x RTX 3090 reference rig - 560GB unified memory, 153 GB/s aggregate bandwidth
Schematic of the NVIDIA Jetson Orin NX 16GB reference rig - 16GB unified memory, 102 GB/s aggregate bandwidth

22 reference configs, drawn in-house. Scroll for more.

Can you run it? - reference rigs

Rig Q4_K_M
Single GTX 1080 Ti (11GB) tight
4x H100 80GB (320GB) fast 1246.6t/s
NVIDIA DGX Station 748GB fast 744.3t/s
8x RTX 3090 rack (192GB) fast 696.8t/s
4x RTX 5090 (128GB) fast 666.9t/s
AMD Instinct MI300X (192GB) fast 495.4t/s
4x RTX 4090 (96GB) fast 375.1t/s
2x RTX 5090 (64GB) fast 333.4t/s
2x RTX 3090 (48GB) fast 174.2t/s
Single RTX 5090 (32GB) fast 166.7t/s
RTX PRO 6000 Blackwell (96GB) fast 166.7t/s
Mac Studio M4 Ultra 192GB fast 110.8t/s
Mac Studio M4 Ultra 512GB fast 110.8t/s
Single RTX 4090 (24GB) fast 93.8t/s
MacBook Pro M5 Max 128GB fast 62.3t/s
Dual EPYC 9004 + 768GB DDR5-4800 fast 42.9t/s
DGX Spark 128GB unified fast 25.4t/s
Ryzen AI Max+ 395 128GB fast 23.8t/s
Jetson AGX Orin 64GB ok 19.1t/s
Epyc + 512GB DDR4-3200 + 2x RTX 3090 ok 19.1t/s
Epyc + 512GB DDR4-2400 + 2x RTX 3090 ok 14.3t/s
NVIDIA Jetson Orin NX 16GB ok 9.5t/s

Fit tiers use the same will-it-run logic as the rig finder. For comfortable fits, the badge reflects decode speed: fast >=20 t/s, ok 8-20 t/s, slow <8 t/s. t/s is a bandwidth estimate, not a measured benchmark.

Download options

Q4_K_M community
5.4GB dl 8.0GB min 16.0GB rec
REC RAM vs largest quant
5.4GB q4 weights + KV 128MB/1k (32 layers x 4 kv x 256); full 262k ctx wants ~40GB, working routing ctx fits 16GB
GGUF on HF →

Or run it in the cloud

No per-token API provider pricing tracked for Clef-flash yet. For flagship list prices, see the calculator.

PRICE HISTORY

Inference cost over time

Data accumulates from the first daily sync - longer ranges populate over time. Prices come from OpenRouter snapshots, not a historical API.

Loading price history...