GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
sahilchachra
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
sahilchachra
Model Tree
mistralai/Leanstral-1.5-119B-A6B
Input Modalities
Output Modalities
Supported Functionality
Composed Mistral3
= DeepSeek-V3 text backbone (MLA + shared-expert MoE: 128 routed experts,
4 active/token, 1 shared) + Pixtral vision. Built by mapping native tensors →
HF names, fusing per-expert weights into the fused MoE layout, and dequantizing
FP8→BF16.
Mistral3(deepseek_v3+pixtral); loads cleanly in transformers.preprocessor_config.json) but not
runtime-validated.yarn and llama_4_scaling; only
YARN is mapped. Long-context (>8k) not guaranteed. Short context unaffected.FP8 e4m3 (per-tensor scale) → BF16 (lossless upcast). Note these weights were
already 8-bit in the source, not full-precision training weights.
Tokenizer is Tekken (tekken.json, via mistral_common); Mistral instruct
chat format (no manual [INST]); base recommends temperature 1.0,
reasoning-parser mistral, context ≤ 200k. Example:
bash
vllm serve sahilchachra/Leanstral-1.5-119B-A6B-BF16 \--tokenizer-mode mistral --tensor-parallel-size 4 \--reasoning-parser mistral --tool-call-parser mistral --enable-auto-tool-choice
For lower memory / single-GPU, prefer the NVFP4 build:
sahilchachra/Leanstral-1.5-119B-A6B-NVFP4.
Files: config.json, weights + index, tekken.json, params.json,
chat_template.jinja.
Derives from mistralai/Leanstral-1.5-119B-A6B; upstream license/terms apply.
Community conversion, not affiliated with Mistral AI.