JEV 27B VL — Hardware Requirements & GPU Compatibility
VisionFunctions
JEV-27B-VL is autotrust's 27-billion-parameter vision decision model, built on Qwen3.8-27B with a LoRA adapter and a decision head. Instead of writing free-form text, its System 1 mode takes a state as text or images together with typed questions (yes/no, pick one of 2 to 256 options, or a 0 to 5 rating) and returns a calibrated probability for each option in one forward pass. The same checkpoint also works as the unmodified Qwen3.8-27B for step-by-step reasoning over images. The card reports first place among 20 vision decision models on the Jev Decision Index 0.3 Vision board, with a Full score of 69.82. The card recommends a single GPU with 80 GB or more for serving; with quantization a 24 GB card is a tight fit. The context length is 262,144 tokens. It is released under the Apache 2.0 license, permitting unrestricted commercial and research use. Published in September 2026, it is a decision-model fine-tune of Qwen3.8-27B and is related in purpose to the Clef and d1 decision models.
1.5M downloads
3.7K likesSpecifications
- Publisher
- autotrust
- Parameters
- 27.8B
- Architecture
- Qwen3_5ForConditionalGeneration
- Context Length
- 262,144 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-09-30
- License
- Apache 2.0
Get Started
HuggingFace
Run in cloud
Fits on RTX 3090 (24 GB) (6 GB headroom) · Q4_K_M
- Generation speed
- ~35 tok/s
- generation speed
- Cost per 1M output tokens
- $1.14
- per 1M output tokens
How Much VRAM Does JEV 27B VL Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| Q2_Kest. | 3.40 | 12.6 GB | 26.4 GB | 11.81 GB | 2-bit quantization with K-quant improvements |
| Q3_K_M | 3.90 | 14.3 GB | 28.2 GB | 13.54 GB | 3-bit medium quantization |
| Q4_K_M | 4.80 | 17.4 GB | 31.3 GB | 16.67 GB | 4-bit medium quantization — most popular sweet spot |
| Q5_K_M | 5.70 | 20.5 GB | 34.4 GB | 19.79 GB | 5-bit medium quantization — good quality/size tradeoff |
| Q6_Kest. | 6.60 | 23.7 GB | 37.5 GB | 22.92 GB | 6-bit quantization, very good quality |
| Q8_0 | 8.00 | 28.5 GB | 42.4 GB | 27.78 GB | 8-bit quantization, near-lossless |
| BF16 | 16.00 | 56.3 GB | 70.2 GB | 55.56 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 JEV 27B VL?
Q4_K_M · 17.4 GBJEV 27B VL (Q4_K_M) requires 17.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 23+ GB is recommended. Using the full 262K context window can add up to 13.9 GB, bringing total usage to 31.3 GB. 8 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti.
Runs great
— Plenty of headroom
Which Devices Can Run JEV 27B VL?
Q4_K_M · 17.4 GB41 devices with unified memory can run JEV 27B VL, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Mini M4 Pro (24 GB).
Runs great
— Plenty of headroomDecent
— Enough memory, may be tightWhere to Download JEV 27B VL
Community quantizations of this model — GGUF for llama.cpp, Ollama, and LM Studio, plus AWQ/MLX variants where available.
Related Models
Frequently Asked Questions
- How much VRAM does JEV 27B VL need?
JEV 27B VL requires 17.4 GB of VRAM at Q4_K_M, or 56.3 GB at BF16. Full 262K context adds up to 13.9 GB (31.3 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 27.8B × 4.8 bits ÷ 8 = 16.7 GB
KV Cache + Overhead ≈ 0.7 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.
KV Cache + Overhead ≈ 14.6 GB (at full 262K context)
VRAM usage by quantization
Q4_K_M17.4 GBQ4_K_M + full context31.3 GB- Can NVIDIA GeForce RTX 4090 run JEV 27B VL?
Yes, at Q6_K (23.7 GB) or lower. Higher quantizations like Q8_0 (28.5 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.
- What's the best quantization for JEV 27B VL?
For JEV 27B VL, Q4_K_M (17.4 GB) offers the best balance of quality and VRAM usage. Q5_K_M (20.5 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 12.6 GB.
VRAM requirement by quantization
Q2_K12.6 GBQ4_K_M ★17.4 GBQ5_K_M20.5 GBQ6_K23.7 GBQ8_028.5 GBBF1656.3 GB★ Recommended — best balance of quality and VRAM usage.
- Can I run JEV 27B VL on a Mac?
Yes — Mac Mini M4 Pro (24 GB) and 22 other Macs can run JEV 27B VL. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 17.4 GB of usable unified memory (RAM minus macOS overhead).
- Can I run JEV 27B VL locally?
Yes — JEV 27B VL can run locally on consumer hardware. At Q4_K_M quantization it needs 17.4 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is JEV 27B VL?
At Q4_K_M, JEV 27B VL can reach ~276 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~38 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 17.4 × 0.65 = ~299 tok/s
Estimated speed at Q4_K_M (17.4 GB)
~299 tok/s~38 tok/s~299 tok/s~276 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of JEV 27B VL?
At Q4_K_M, the download is about 16.67 GB. The full-precision BF16 version is 55.56 GB. The smallest option (Q2_K) is 11.81 GB.
- Which GPUs can run JEV 27B VL?
8 consumer GPUs can run JEV 27B VL at Q4_K_M (17.4 GB). Top options include NVIDIA GeForce RTX 5090, AMD Radeon RX 7900 XT, AMD Radeon RX 7900 XTX. 1 GPU have plenty of headroom for comfortable inference.
- Which devices can run JEV 27B VL?
41 devices with unified memory can run JEV 27B VL at Q4_K_M (17.4 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.