Clef Omni — Hardware Requirements & GPU Compatibility
VisionFunctions
Clef-Omni is Cloudflare's multimodal decision model, post-trained from Qwen3-Omni-30B-A3B-Instruct, with about 35 billion parameters in the full checkpoint and roughly 3 billion active in the mixture-of-experts backbone. It does not generate free-form text. It reads a state as text, JSON, images, audio or video together with a schema of typed questions, and returns a probability for every allowed option in one forward pass, using a small joint schema head on the backbone's hidden states. The base model's speech-output weights are included but unused. At this size it fits a 24 GB card only at heavy quantization, and the custom head requires the repository's own loading code. The backbone configuration lists a context length of 65,536 tokens, and the card's encoder defaults to 64,000. It is released under the Apache 2.0 license, permitting unrestricted commercial and research use. Published in October 2026, it is the audio-capable companion to the dense Clef and Clef-Flash models.
2 downloads
29 likesSpecifications
- Publisher
- Cloudflare
- Parameters
- 35.3B
- Architecture
- Qwen3OmniMoeForConditionalGeneration
- Release Date
- 2026-10-09
- License
- Apache 2.0
Get Started
HuggingFace
Run in cloud
Fits on RTX A6000 (48 GB) (24 GB headroom) · Q4_K_M
- Generation speed
- ~143 tok/s
- generation speed
- Cost per 1M output tokens
- $0.64
- per 1M output tokens
How Much VRAM Does Clef Omni Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| Q2_Kest. | 3.40 | 16.5 GB | — | 14.99 GB | 2-bit quantization with K-quant improvements |
| Q3_K_Mest. | 3.90 | 18.9 GB | — | 17.19 GB | 3-bit medium quantization |
| Q4_K_Mest. | 4.80 | 23.3 GB | — | 21.16 GB | 4-bit medium quantization — most popular sweet spot |
| Q5_K_Mest. | 5.70 | 27.6 GB | — | 25.12 GB | 5-bit medium quantization — good quality/size tradeoff |
| Q6_Kest. | 6.60 | 32 GB | — | 29.09 GB | 6-bit quantization, very good quality |
| Q8_0est. | 8.00 | 38.8 GB | — | 35.26 GB | 8-bit quantization, near-lossless |
| BF16est. | 16.00 | 77.6 GB | — | 70.52 GB | Brain floating point 16 — preferred for training |
est.= calculated VRAM estimate; no published GGUF file found for that quantization yet. Other rows are verified against real community uploads.
Which GPUs Can Run Clef Omni?
Q4_K_M · 23.3 GBClef Omni (Q4_K_M) requires 23.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 31+ GB is recommended. 7 GPUs can run it, including NVIDIA GeForce RTX 5090.
All compatible consumer-level GPUs are running near their VRAM limit. You may also want to consider professional GPUs (e.g., NVIDIA A100, H100) which offer significantly more VRAM. For more headroom and better throughput, consider a multi-GPU configuration with tensor parallelism (supported by tools like vLLM, llama.cpp, or text-generation-inference).
Which Devices Can Run Clef Omni?
Q4_K_M · 23.3 GB35 devices with unified memory can run Clef Omni, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Studio M4 Max (36 GB).
Runs great
— Plenty of headroomDecent
— Enough memory, may be tightRelated Models
Frequently Asked Questions
- How much VRAM does Clef Omni need?
Clef Omni requires 23.3 GB of VRAM at Q4_K_M, or 77.6 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 35.3B × 4.8 bits ÷ 8 = 21.2 GB
KV Cache + Overhead ≈ 2.1 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
VRAM usage by quantization
Q4_K_M23.3 GB- Can NVIDIA GeForce RTX 4090 run Clef Omni?
Yes, at Q4_K_M (23.3 GB) or lower. Higher quantizations like Q5_K_M (27.6 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
- What's the best quantization for Clef Omni?
For Clef Omni, Q4_K_M (23.3 GB) offers the best balance of quality and VRAM usage. Q5_K_M (27.6 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 16.5 GB.
VRAM requirement by quantization
Q2_K16.5 GBQ4_K_M ★23.3 GBQ5_K_M27.6 GBQ6_K32.0 GBQ8_038.8 GBBF1677.6 GB★ Recommended — best balance of quality and VRAM usage.
- Can I run Clef Omni on a Mac?
Yes, but only at lower quantizations. The smallest Mac that can run Clef Omni is Mac Mini M4 Pro (24 GB) at Q2_K; 23 of the 39 Macs we list can run it at some quantization. For Q4_K_M (23.3 GB) you need a Mac with more unified memory.
- Can I run Clef Omni locally?
Yes — Clef Omni can run locally on consumer hardware. At Q4_K_M quantization it needs 23.3 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Clef Omni?
At Q4_K_M, Clef Omni can reach ~100 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~168 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: only the active experts are read per token, plus a fixed per-layer routing cost.
Example: NVIDIA B200 → 2.2 GB active ÷ (8000 × 0.65) = 0.42 ms, plus 48 layers × 0.055 ms = 2.64 ms, so 1000 ÷ 3.06 ms = ~327 tok/s
Estimated speed at Q4_K_M (23.3 GB)
~327 tok/s~168 tok/s~327 tok/s~301 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Clef Omni?
At Q4_K_M, the download is about 21.16 GB. The full-precision BF16 version is 70.52 GB. The smallest option (Q2_K) is 14.99 GB.
- Which GPUs can run Clef Omni?
7 consumer GPUs can run Clef Omni at Q4_K_M (23.3 GB). Top options include AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090.
- Which devices can run Clef Omni?
35 devices with unified memory can run Clef Omni at Q4_K_M (23.3 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Asus ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB), Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.