Mixtral 8x22B

CONTEXT SCALING: null is a VERIFIED value here, checked against config rather than left blank. config.json (mistralai/Mixtral-8x22B-Instruct-v0.1) declares rope_scaling null, rope_theta 1000000, max_position_embeddings 65536. Verified null. PARAMS: 140.6B is the safetensors count; Mistral's post says 141B. ACTIVE: 39B is Mistral's published figure, not derived. KV: 56 * 8 * 128 * 2 * 2 bytes = 229,376 bytes/token = 224 KiB/token, well under half of Llama 3.3 70B's 320 KiB/token despite being twice the weights and 39B active versus 70B. This is the cleanest KV-vs-active divergence in the open-weight set: 39B of active compute with 224 KiB/token of KV. Same 2-of-8 expert ratio as 8x7B but each expert is much wider (intermediate 16384 vs 14336 against a 6144 vs 4096 hidden size), which is why active params scale to 39B. 64k context per the vendor post and config. SERVICEABILITY JOIN (added 2026-10-03; the architecture x engine matrix this record was missing, with a citation per cell in docs/13-model-compatibility.md): FAMILY=dense-gqa. REQUIRED KERNEL CLASS: ordinary causal attention plus GEMM/GEMV. This is the baseline every engine in this repo implements, so the join here is a floor rather than a filter: no engine record in data/engines/ states that dense GQA is unsupported. SUPPORTED BY (cited): vLLM LlamaForCausalLM / Qwen2 / Qwen3 / Mistral / Phi3ForCausalLM rows [[mdl2-vllm-supported-models-2026-10]]; TensorRT-LLM LlamaForCausalLM, Qwen2ForCausalLM, Qwen3ForCausalLM, Phi3ForCausalLM rows [[mdl2-trtllm-supported-models]]; SGLang generative table (Llama, Qwen, Mistral families) [[mdl2-sglang-generative-models]]; llama.cpp LLM_ARCH_LLAMA [[pap-arch-llamacpp-arch-registry]]. HARD CONSTRAINTS THAT DO APPLY even to dense GQA, and they are model-shape rather than architecture-class: tensor-parallel size must divide the attention head count or the engine refuses to start [[mgpu-tp-size-must-divide-attention-heads]] (affects vllm and sglang); and a GQA KV cache is the thing paged-attention machinery exists for, so an engine with no paged KV (llama.cpp's slot-based pool [[llama-cpp]], exllamav2 [[exllamav2]]) loses the concurrency lever rather than the ability to run. ONE HARDWARE-DEPENDENT KERNEL GATE THAT IS NOT ABOUT GQA: on RDNA, vLLM's custom paged-attention kernel is gated on head_size == 128 AND block_size == 16, so a dense GQA model with head_dim 256 falls to the Triton 2D fallback and decode falls off a cliff - measured 12.1 tok/s at 518 tokens of context to 4.2 tok/s at 32k, with kernel_paged_attention_2d growing 28.3x [[rdna-triton-paged-attn-decode-cliff]]. Check head_dim before blaming the card.