GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
darrellbest
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
darrellbest
Model Tree
darrellbest/Qwen3-0.6B-Heretic
Input Modalities
Output Modalities
Supported Functionality
| Part | Precision |
|---|---|
| All attention and MLP linear layers (196) | NVFP4, 16-value groups, FP8 scales |
Embeddings (tied to lm_head), norms | bf16 (unchanged) |
The bf16 embedding table is about 0.31 GB of the 0.57 GB file; the quantized layers themselves are small.
Made with llm-compressor 0.13.0 (scheme="NVFP4"), calibrated on
64 harmless chat prompts. Format: compressed-tensors, nvfp4-pack-quantized.
sh
vllm serve darrellbest/Qwen3-0.6B-Heretic-NVFP4Loaded in vLLM 0.30.0 on an RTX PRO 6000 Blackwell: ordinary prompts answered correctly (wording drifts a little from
the bf16 model, as expected at 4 bits), and thinking mode reasons to the right answer on arithmetic and closes its
<think> block. Throughput: ~310 tok/s single-stream, ~6,600 tok/s aggregate at batch 32 (512-token generations,
measured on a shared GPU). At 0.6B, NVFP4 buys size, not speed.
3/100 refusals against 54/100 for the original, KL divergence 0.0027. Method and parameters are on the parent card. The 4-bit weights were not re-measured for refusals.
Reduced safety guardrails by design. You are responsible for what you do with it.
| Repository | Format | Size | Use it with |
|---|---|---|---|
| Qwen3-0.6B-Heretic | bf16 safetensors | 1.19 GB | transformers, vLLM, SGLang |
| Qwen3-0.6B-Heretic-GGUF | GGUF BF16 / Q8_0 / Q4_K_M | 1.20 / 0.64 / 0.40 GB | llama.cpp, Ollama |
| Qwen3-0.6B-Heretic-FP8 | FP8 W8A8, compressed-tensors | 0.76 GB | vLLM |
| Qwen3-0.6B-Heretic-NVFP4 | NVFP4, compressed-tensors | 0.57 GB | vLLM on Blackwell |