Open-weights models, addressed by content and served by the network.

Google's jointly-trained Multi-Token Prediction head for Gemma 4 E2B. Pair with the E2B target via `--spec-type draft-mtp`.

IBM Granite 4.0 350M — ultra-compact for edge deployment

Tiny Gemma model for ultra-lightweight on-device inference

Google's jointly-trained Multi-Token Prediction head for Gemma 4 E4B. Pair with the E4B target via `--spec-type draft-mtp`.

Compact model optimized for edge deployment

Speculative drafter for Mistral Small 3.1/3.2 — vocab-matched, 6-language fine-tune.

Compact multilingual model for efficient on-device inference

Qwen 3.5 0.8B with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth measures ~1.5-2× speedup over the non-MTP baseline.

Google's jointly-trained Multi-Token Prediction head for Gemma 4 12B. Pair with the 12B target via `--spec-type draft-mtp`.

IBM Granite 4.0 1B — compact enterprise model

Google's compact instruction-tuned model

Compact SmolLM2 for on-device AI

Versatile model for various language tasks

Google's jointly-trained Multi-Token Prediction head for the Gemma 4 26B-A4B Mixture-of-Experts target. Pair via `--spec-type draft-mtp`; Unsloth measures ~1.15–1.2× speedup on MoE targets vs ~1.4–2.2× on dense.

Efficient small model for chat and text generation

Qwen 3.5 2B with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth measures ~1.5-2× speedup over the non-MTP baseline.

Google's jointly-trained Multi-Token Prediction head for Gemma 4 31B. Pair with the 31B target via `--spec-type draft-mtp`.

SmolLM3 — 11T tokens, dual-mode reasoning

Compact Ministral 3 for lightweight tasks

Extended context Gemma model for chat applications

Compact Phi-4 Mini with 128K context

Compact Phi-4 Mini fine-tuned for reasoning tasks

Well-balanced model for production use

Qwen 3.5 4B with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth measures ~1.5-2× speedup over the non-MTP baseline.

Mid-size model with strong reasoning and coding performance

Hybrid Mamba-2 + Attention edge model, 256K context

Google's compact Gemma 4 multimodal model (text + image, 128K context). MTP-enabled — pairs with `gemma4-e2b-mtp-draft` for 1.5–2.2× throughput.

Quantization-Aware-Trained Gemma 4 E2B. Higher quality than naive Q4 at the same size. MTP-enabled via `gemma4-e2b-mtp-draft`.

IBM Granite 4.0 H-Tiny — hybrid Mamba/Transformer architecture

Mistral AI's classic 7B instruction model

Versatile Ministral 3 for general-purpose tasks

Extended context model for long-form tasks

Google's efficient Gemma 4 multimodal model (text + image, 128K context). MTP-enabled — pairs with `gemma4-e4b-mtp-draft` for 1.5–2.2× throughput.

Quantization-Aware-Trained Gemma 4 E4B. Higher quality than naive Q4 at the same size. MTP-enabled via `gemma4-e4b-mtp-draft`.

Qwen 3.5 9B with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth measures ~1.5-2× speedup over the non-MTP baseline.

High-performance model for complex language understanding

Zhipu AI GLM-4 9B instruction-tuned, 128K context

High-performance instruction-tuned model from Google

Extended-context Mistral model built with NVIDIA

Google's mid-tier dense Gemma 4 model (128K context). MTP-enabled — pairs with `gemma4-12b-mtp-draft` for 1.5–2.2× throughput on the same hardware (Unsloth: 52 → 162 t/s at Q4 on a 4090).

Quantization-Aware-Trained Gemma 4 12B. Higher quality than naive Q4 at the same size. MTP-enabled via `gemma4-12b-mtp-draft`.

High-performance Ministral 3 for complex reasoning

Microsoft Phi-4 — strong reasoning at 14B

Premium model with extended context support

Phi-4 fine-tuned for chain-of-thought reasoning

OpenAI GPT-OSS 20B — open-weights release, native MXFP4

Mistral Small 3.1 — improved reasoning over 3.0 baseline

Mistral Small 3.2 — latest 3-series point release

Mistral's latest Small 3.2 model for demanding workloads

Google's largest Gemma model with exceptional capabilities

Flagship Qwen 3.5 model

Qwen 3.6 27B — flagship dense model with 128K context

Qwen 3.5 27B with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth measures ~1.5-2× speedup over the non-MTP baseline.

Qwen 3.6 27B with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth: 160 t/s on RTX 6000.

Diffusion-generation Gemma 4 26B-A4B. Generates 256-token canvases by parallel denoising rather than autoregressive sampling. Unsloth: 2000+ t/s on RTX 6000. Requires Unsloth Studio or llama.cpp PR #24423+.

Gemma 4 Mixture-of-Experts: 26B total params, 4B active per token (128K context). MTP-enabled — pairs with `gemma4-26b-a4b-mtp-draft`; expect ~1.15–1.2× speedup on MoE targets per Unsloth.

Quantization-Aware-Trained Gemma 4 26B-A4B MoE. Unsloth measures 85.6% MMLU top-1 vs 70.2% on naive Q4 (+15.4 points). MTP-enabled via `gemma4-26b-a4b-mtp-draft`.

Mixture-of-Experts with 3B active params for efficient scaling

Code-focused MoE — 30B total, 3B active, 256K context

IBM Granite 4.0 H-Small — 32B hybrid for long-context enterprise

Google's largest dense Gemma 4 model (128K context). MTP-enabled — pairs with `gemma4-31b-mtp-draft` for ~2× throughput at 101 t/s on consumer GPUs (Unsloth benchmark).

Top-tier model with 128K context window

Quantization-Aware-Trained Gemma 4 31B. Higher quality than naive Q4 at the same size. MTP-enabled via `gemma4-31b-mtp-draft`.

Moonshot AI Kimi K2 MoE — 1T total, 32B active, 128K context

Qwen 3.6 MoE — 35B total, ~3B active per token

Qwen 3.6 35B-A3B MoE with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth: 240 t/s on RTX 6000.

Mixture-of-Experts with only 3B active params — fast inference at 35B quality

Qwen 3.5 35B-A3B (MoE) with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth measures ~1.5-2× speedup over the non-MTP baseline.

Hybrid Mamba-2 MoE — 30B total, 3.5B active, 128K context

OpenAI GPT-OSS 120B — open-weights release, native MXFP4

Qwen 3.5 large MoE — 122B total, 10B active per token. Replica-routed on high-VRAM provider tiers only; Unsloth ships an MTP variant in `unsloth/Qwen3.5-122B-A10B-MTP-GGUF` for compatible runtimes.

Qwen 3.5 122B-A10B (MoE) with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth measures ~1.5-2× speedup over the non-MTP baseline.

MiniMax M2.7 — current frontier MiniMax MoE; Lightning Attention with 1M context. Unsloth dynamic UD-Q4_K_XL GGUF (sharded).

DeepSeek V4 Flash — 284B total / 13B active MoE; 1M context. Hybrid Compressed Sparse Attention (CSA) + Heavily Compressed Attention (HCA). Cost-effective frontier variant. Pre-trained on 32T tokens. MTP head shipped natively.

MiniMax M3 — ~428B total / ~23B active MoE with native multimodal training. MiniMax Sparse Attention (MSA) delivers 9× prefill and 15× decode speedups vs M2 at 1M context. Note: GGUF builds currently fall back to dense attention; sparse attention not yet supported in llama.cpp.

Qwen 3.5 frontier MoE — 397B total, 17B active per token. Multi-GPU replicas only; Unsloth ships an MTP variant in `unsloth/Qwen3.5-397B-A17B-MTP-GGUF` for compatible runtimes.

Qwen 3.5 397B-A17B (MoE) with built-in Multi-Token-Prediction head. Single-file MTP GGUF — no separate drafter needed. Unsloth measures ~1.5-2× speedup over the non-MTP baseline.

DeepSeek V3 MoE — 685B total, 37B active, 128K context. Native Multi-Token-Prediction head (n=4, ~80% accept rate, ~1.8× decode speedup per DeepSeek tech report). Retired by upstream after 2026-07-24 in favor of DeepSeek V4.

Z.ai GLM-5 — 744B total parameter MoE trained on 28.5T tokens; best-in-class open-source performance on reasoning, coding, and agentic tasks (2026-04). Unsloth dynamic UD-Q4_K_XL GGUF (sharded).

Z.ai GLM-5.1 — next-generation flagship for agentic engineering, class-leading on SWE-Bench Pro; 744B total / 40B active; 200K context. `glm_moe_dsa` architecture with Dynamic Sparse Attention.

Z.ai GLM-5.2 — 753B total parameter MoE flagship with solid 1M-token context and IndexShare sparse-attention (2.9× per-token FLOP reduction at 1M). Improved Multi-Token-Prediction layer increases speculative-decoding accept rate by ~20% over GLM-5.1.

Moonshot AI Kimi K2.5 — 1T total / 32B active MoE; image input support; 256K context. Predecessor to K2.6's hybrid-thinking variant.

Moonshot AI Kimi K2.7 Code — code-focused refresh of the K2 series. 1T total / 32B active; 256K context; recent updates target tool-call accuracy on long-horizon coding tasks.

Moonshot AI Kimi K2.6 hybrid-thinking MoE — 1T total params, 256K context. Replica-routed on B200-class infrastructure; Unsloth measures >40 t/s on B200. Recommended `UD-Q2_K_XL` (350GB) for size/quality balance.