Advanced tool
Answers: Which models and configurations RigFact knows about, and their listing status.
Does not answer: Whether RigFact tested them, whether they fit your GPU, or whether they suit your workload.
Model catalog
Every model RigFact lists, with its status, its publisher source and its evidence state. Catalog presence means the model exists and is discoverable — it does NOT mean RigFact tested it, that it fits your GPU, or that we recommend it.
Seeing a model here does not mean RigFact tested it, that it is compatible with your GPU, that it is recommended, or that it suits your workload.
Current means relevant and available as of 2026 — not necessarily released in 2026. Older releases stay listed while they remain current choices, and a recent release date alone proves nothing. A release date is provenance, not a currency verdict. A primary publisher source confirms the model exists and what it is. It is not evidence that RigFact tested it, that it fits your GPU, how it performs, or that a Recipe exists. Freshness reviewed 2026-10-03.
What this calculator covers
Only models whose architecture parameters are source-verified.The calculator only estimates models whose architecture parameters are source-verified. A current catalog model appears here once its architecture is verified — it is not missing because it is old, and its absence is not a fit or quality verdict.
17 current 2026 catalog models are not estimated here yet — their architecture parameters are not source-verified, so no estimate is invented for them: Qwen3.8 27B · Qwen3.6 27B · Qwen3.6 35B A3B · Muse Glimmer 30B · Qwen3-Coder 30B A3B · Gemma 4 26B MoE · gpt-oss 20B · Mistral Small 3.2 24B · DeepSeek-R1-Distill-Qwen 32B · GLM-4.7-Flash · GLM-5.3 · Kimi K3 · DeepSeek V4-Pro · Qwen3.5 397B · Llama 4 Scout · gpt-oss 120B · Mistral Medium 3.5. They stay discoverable under Models relevant to local AI in 2026 and in the model catalog. Absence from the calculator is not a currency, fit or quality verdict.
Tested RigFact Recipes are a third, separate thing: executed configurations with captured evidence. A calculator estimate is not a Recipe, and a catalog entry is neither.
Current (17)
Relevant and available as of 2026 — not necessarily released in 2026.
Qwen3.8 27B
16–32 GB class (sourced)Qwen3.8 · 27B · Released 2026-08-14 (Alibaba / Qwen)
Tested by RigFact — first-party Recipe
7 first-party Recipe(s) — see Hardware First.
- Publisher source:
- Alibaba / Qwen Team — Qwen/Qwen3.8-27B
- Calculator:
- architecture not source-verified — not estimated
Qwen3.6 27B
16–32 GB class (sourced)Qwen3.6 · 27B · Released 2026-04 (Alibaba / Qwen)
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Alibaba / Qwen Team — Qwen/Qwen3.6-27B
- Calculator:
- architecture not source-verified — not estimated
Qwen3.6 35B A3B
16–32 GB class (sourced)Qwen3.6 · 35B-A3B · Qwen3.6 generation (2026)
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Alibaba / Qwen Team — Qwen/Qwen3.6-35B-A3B
- Calculator:
- architecture not source-verified — not estimated
Muse Glimmer 30B
16–32 GB class (sourced)Muse Glimmer · 30B · Released 2026-08-10 (Meta), Apache 2.0
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Meta Superintelligence Lab (Meta Inc.) — meta-models/Muse-Glimmer-30B
- Calculator:
- architecture not source-verified — not estimated
Qwen3-Coder 30B A3B
16–32 GB class (sourced)Qwen3-Coder · 30B-A3B · Released 2025-07 (Alibaba / Qwen), Apache 2.0, 262 144-token context
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Alibaba / Qwen Team — Qwen/Qwen3-Coder-30B-A3B-Instruct
- Calculator:
- architecture not source-verified — not estimated
Gemma 4 26B MoE
Sub-16 GB class (sourced)Gemma 4 · 26B MoE · Released 2026-04-02 (Google DeepMind), Apache 2.0
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Google DeepMind
- Calculator:
- architecture not source-verified — not estimated
gpt-oss 20B
Sub-16 GB class (sourced)gpt-oss · 20b · Apache 2.0 open-weight release (OpenAI)
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- OpenAI — openai/gpt-oss-20b
- Calculator:
- architecture not source-verified — not estimated
Mistral Small 3.2 24B
16–32 GB class (sourced)Mistral Small · 3.2 24B · Mistral Small 3.2 generation, Apache 2.0
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Mistral AI — mistralai/Mistral-Small-3.2-24B-Instruct-2506
- Calculator:
- architecture not source-verified — not estimated
DeepSeek-R1-Distill-Qwen 32B
16–32 GB class (sourced)DeepSeek-R1-Distill · Qwen 32B · DeepSeek R1 distill line, MIT licence
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- DeepSeek — deepseek-ai/DeepSeek-R1-Distill-Qwen-32B
- Calculator:
- architecture not source-verified — not estimated
GLM-4.7-Flash
16–32 GB class (sourced)GLM · 4.7 Flash (30B MoE) · Zhipu AI, 200K context, listed as open weights
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Z.ai / Zhipu AI — zai-org/GLM-4.7-Flash
- Calculator:
- architecture not source-verified — not estimated
GLM-5.3
Server-class / multi-GPU (sourced)GLM · 5.3 · Released 2026-08-18 (Z.ai); GLM-5.2 generation was a 753B MoE
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Z.ai / Zhipu AI — zai-org/GLM-5.3
- Calculator:
- architecture not source-verified — not estimated
Kimi K3
Server-class / multi-GPU (sourced)Kimi · K3 · Moonshot AI; ~2.8T total parameters
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Moonshot AI — moonshotai/Kimi-K3
- Calculator:
- architecture not source-verified — not estimated
DeepSeek V4-Pro
Server-class / multi-GPU (sourced)DeepSeek · V4-Pro · DeepSeek V4 line; near 1.6T total parameters
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- DeepSeek — deepseek-ai/DeepSeek-V4-Pro
- Calculator:
- architecture not source-verified — not estimated
Qwen3.5 397B
Server-class / multi-GPU (sourced)Qwen3.5 · 397B · Alibaba / Qwen3.5 line (server-class scale)
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Alibaba / Qwen Team — Qwen/Qwen3.5-397B-A17B
- Calculator:
- architecture not source-verified — not estimated
Llama 4 Scout
Server-class / multi-GPU (sourced)Llama 4 · Scout · Meta Llama 4 line
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Meta (Meta Llama) — meta-llama/Llama-4-Scout-17B-16E-Instruct
- Calculator:
- architecture not source-verified — not estimated
gpt-oss 120B
Server-class / multi-GPU (sourced)gpt-oss · 120b · OpenAI open-weight line
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- OpenAI — openai/gpt-oss-120b
- Calculator:
- architecture not source-verified — not estimated
Mistral Medium 3.5
Server-class / multi-GPU (sourced)Mistral Medium · 3.5 (128B) · Mistral AI — official repository mistralai/Mistral-Medium-3.5-128B
Not tested by RigFact
Catalog presence is not evidence.
- Publisher source:
- Mistral AI — mistralai/Mistral-Medium-3.5-128B
- Calculator:
- architecture not source-verified — not estimated
Publisher-deprecated (1)
Kept discoverable for history and comparison. This is a publisher lifecycle statement — not a RigFact quality judgement, and not an age-based demotion.
Devstral Small 2 24B
LEGACYDevstral Small · 2 (24B)
Not tested by RigFact
The publisher documentation marks this model DEPRECATED and names Mistral Medium 3.5 as the replacement. This is a sourced publisher status, not a RigFact judgement, and it does not change any evidence state.
- Status reason:
- PUBLISHER_DEPRECATED — the publisher marks this model deprecated and names a replacement. Sourced publisher lifecycle status, not a RigFact quality judgement.
- Publisher source:
- Mistral AI — mistralai/Devstral-Small-2-24B-Instruct-2512
Preserved legacy (18)
Pre-2026 entries kept for history and comparison. Never erased, never silently promoted to current. A release date is provenance, not a currency verdict.
Llama 3.3 70B
Not tested by RigFact
Qwen 2.5 72B
Not tested by RigFact
DeepSeek R1 70B
Not tested by RigFact
Qwen 2.5 32B
Not tested by RigFact
Llama 3.1 8B
Not tested by RigFact
Mistral 8x22B (MoE)
Not tested by RigFact
Gemma 3 27B
Not tested by RigFact
Mistral Large 2 123B
Not tested by RigFact
DeepSeek V3 671B (MoE)
Not tested by RigFact
Command R+ 104B
Not tested by RigFact
Qwen 2.5 7B
Not tested by RigFact
Mistral Small 22B
Not tested by RigFact
Phi-3 Medium 14B
Not tested by RigFact
Mistral 7B
Not tested by RigFact
Gemma 2 7B
Not tested by RigFact
Phi-3 Mini 4B
Not tested by RigFact
Llama 3.2 3B
Not tested by RigFact
Llama 3.2 1B
Not tested by RigFact