What can it run

Model card lists 0.8B parameters and a native 262,144-token context window.

Parameter count from Qwen3.5-0.8B model card, fetched 2026-09-15

Model card lists 1.17B parameters and a 32,768-token context window.

Parameter count from LFM2.5-1.2B-Thinking model card, fetched 2026-09-15

Model card lists 2B parameters and a native 262,144-token context window.

Parameter count from Qwen3.5-2B model card, fetched 2026-09-15

Model card calls it a 3B parameter model; the current config supports 65,536 tokens and the card documents YaRN extensions for 128K and 256K.

Parameter count from SmolLM3-3B model card, fetched 2026-09-15

Model card lists 4B parameters and a native 262,144-token context window.

Parameter count from Qwen3.5-4B model card, fetched 2026-09-15

Google lists 4.5B effective parameters and 8B total with embeddings. Sizing uses all 8B resident weights, not the effective count.

Parameter count from Gemma 4 E4B IT model card, fetched 2026-09-15

Model card: "Number of Parameters: 8.2B". Context 32,768 natively, 131,072 with YaRN.

Parameter count from Qwen3-8B model card, fetched 2026-08-16

Model card lists 9B parameters and a native 262,144-token context window.

Parameter count from Qwen3.5-9B model card, fetched 2026-09-15

Google lists 11.95B total parameters and a 256K context window for the encoder-free multimodal model.

Parameter count from Gemma 4 12B IT model card, fetched 2026-09-15

OpenAI: "21B parameters with 3.6B active parameters"; card says it can run within 16GB of memory.

Parameter count from gpt-oss-120b model card, fetched 2026-08-16

Model card lists 24B params. Mistral does not state a context window on the card.

Parameter count from Mistral Small 3.2 24B Instruct model card, fetched 2026-08-16

Mixture of experts: Google lists 25.2B total parameters and 3.8B active, with a 256K context window.

Parameter count from Gemma 4 26B-A4B IT model card, fetched 2026-09-15

Model card lists 27B params, with a 128K input context window.

Parameter count from Gemma 3 27B IT model card, fetched 2026-08-16

Model card lists 27B parameters and a native 262,144-token context window.

Parameter count from Qwen3.5-27B model card, fetched 2026-09-15

Model card lists 27B parameters and a native 262,144-token context window, extensible to 1,000,000 tokens.

Parameter count from Qwen3.8-27B model card, fetched 2026-09-15

Google lists 30.7B total parameters and a 256K context window for the dense model.

Parameter count from Gemma 4 31B IT model card, fetched 2026-09-15

Mixture of experts: model card lists 35B total parameters and 3B active, with a native 262,144-token context window.

Parameter count from Qwen3.5-35B-A3B model card, fetched 2026-09-15

Card table lists 70B; the computed safetensors count reads 71B. We use the nominal 70B.

Parameter count from Llama-3.3-70B-Instruct model card, fetched 2026-08-16

OpenAI: "117B parameters with 5.1B active parameters"; designed to fit a single 80GB GPU using MXFP4.

Parameter count from gpt-oss-120b model card, fetched 2026-08-16

Mixture of experts: model card lists 122B total parameters and 10B active, with a native 262,144-token context window.

Parameter count from Qwen3.5-122B-A10B model card, fetched 2026-09-15

Mixture of experts: "235B in total and 22B activated". All weights must still be resident.

Parameter count from Qwen3-235B-A22B model card, fetched 2026-08-16

Mixture of experts: the DeepSeek-V4 report lists 284B total parameters and 13B active, with a one-million-token context window.

Parameter count from DeepSeek-V4-Flash-0731 model card, fetched 2026-09-15, DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence, fetched 2026-09-15

"671B total parameters" with "37B activated for each token". The HF upload totals 685B including the MTP module.

Parameter count from DeepSeek-V3 model card, fetched 2026-08-16

Mixture of experts: the official GLM-5 repository lists GLM-5.1 as 744B total parameters with 40B active.

Parameter count from GLM-5.1 model card, fetched 2026-09-15, GLM-5 official repository, fetched 2026-09-15

Mixture of experts: card table reads Total Parameters 2.8T, Activated Parameters 104B, 896 experts with 16 selected per token, 1,048,576 context. All weights must be resident, so nothing we track fits it at any quantization.

Parameter count from Kimi-K3 model card, fetched 2026-08-16