# Best open-weight LLMs: how to choose

Newest frontier open-weight model: Muse Spark Open. Largest context window: GLM-5.3. Lowest API price, if you do not self-host: MiniMax M3.1 Flash. Sized for your own hardware: Nemotron 3.5 Bolt E4B (4GB). "Open weights" means the weights are downloadable; licence terms differ, so read the licence before commercial use. We have not run these models.

Updated: 2026-10-29 · Canonical: https://theagenticweekly.com/best/open-weight-llms · Evidence: Reported (not run by us)

## Picks, and the fact behind each

| Pick | Model | Why | Source |
|---|---|---|---|
| Newest frontier open-weight model | [Muse Spark Open](https://theagenticweekly.com/models/muse-spark-open) | Released Oct 15, 2026 under Muse Model License (limited commercial terms); 816B parameters (42B active). | [source](https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B) |
| Largest context window | [GLM-5.3](https://theagenticweekly.com/models/glm-5-3) | 1,048,576 tokens, the largest published among recent open-weight models here. | [source](https://docs.z.ai/guides/llm/glm-5.3) |
| Lowest API price, if you do not self-host | [MiniMax M3.1 Flash](https://theagenticweekly.com/models/minimax-m3-1-flash) | $0.15 in / $0.6 out per 1M tokens on the vendor API. | [source](https://www.minimax.io/blog/minimax-m31f) |
| Sized for your own hardware | [Nemotron 3.5 Bolt E4B (4GB)](https://theagenticweekly.com/models/nemotron-3-5-bolt-e4b) | Newest model here that we file under "runs on your own hardware" (8B parameters); OpenMDW-1.1. | [source](https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB) |

## All models in scope

Every language model in our data with downloadable weights, newest first.

| Model | Vendor | Licence | Context | Price (listed) | Released | Status |
|---|---|---|---|---|---|---|
| [Nemotron 3.5 Bolt E4B (4GB)](https://theagenticweekly.com/models/nemotron-3-5-bolt-e4b) | NVIDIA | OpenMDW-1.1 ([source](https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB)) | 64K tokens ([source](https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB)) | Not published | Oct 23, 2026 ([source](https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Bolt-E4B-4GB)) | Generally available |
| [MiniMax M3.1 Flash](https://theagenticweekly.com/models/minimax-m3-1-flash) | MiniMax | MiniMax Community License ([source](https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash/blob/main/LICENSE)) | 1M tokens ([source](https://huggingface.co/MiniMaxAI/MiniMax-M3.1-Flash)) | $0.15 in / $0.6 out per 1M tokens ([source](https://www.minimax.io/blog/minimax-m31f)) | Oct 19, 2026 ([source](https://www.minimax.io/blog/minimax-m31f)) | Generally available |
| [DeepSeek-V4.1-Flash-1017](https://theagenticweekly.com/models/deepseek-v4-1-flash-1017) | DeepSeek | MIT ([source](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017)) | 1M tokens ([source](https://api-docs.deepseek.com/zh-cn/quick_start/pricing)) | ¥1 in / ¥4 out per 1M tokens; peak: ¥2 / ¥8 ([source](https://api-docs.deepseek.com/zh-cn/quick_start/pricing)) | Oct 17, 2026 ([source](https://api-docs.deepseek.com/news/news261017/)) | Generally available |
| [Muse Spark Open](https://theagenticweekly.com/models/muse-spark-open) | Meta | Muse Model License (limited commercial terms) ([source](https://huggingface.co/meta-models/Muse-Spark-Open-816B-A42B)) | 1M tokens ([source](https://research.meta.ai/blog/muse-spark-open)) | Not published | Oct 15, 2026 ([source](https://research.meta.ai/blog/muse-spark-open)) | Generally available |
| [Mistral Large 4](https://theagenticweekly.com/models/mistral-large-4) | Mistral AI | Mistral Ecosystem Open-Weight License (MIT derivative with limits on model-as-a-service) ([source](https://huggingface.co/blog/mistral-large4)) | 1M tokens ([source](https://docs.mistral.ai/models/mistral-large-4)) | $1.36 in / $4.18 out per 1M tokens ([source](https://docs.mistral.ai/models/mistral-large-4)) | Oct 6, 2026 ([source](https://mistral.ai/news/mistral-large-4/)) | Public preview |
| [DeepSeek-V4.1-Flash](https://theagenticweekly.com/models/deepseek-v4-1-flash) | DeepSeek | MIT ([source](https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017)) | 1M tokens ([source](https://api-docs.deepseek.com/zh-cn/quick_start/pricing)) | ¥1 in / ¥4 out per 1M tokens; peak: ¥2 / ¥8 ([source](https://api-docs.deepseek.com/zh-cn/quick_start/pricing)) | Sep 10, 2026 ([source](https://api-docs.deepseek.com/updates/)) | Generally available |
| [North Small Translate](https://theagenticweekly.com/models/north-small-translate) | Cohere | CC BY-NC 4.0 ([source](https://docs.cohere.com/changelog)) | 16K tokens ([source](https://docs.cohere.com/changelog)) | Not published | Sep 9, 2026 ([source](https://docs.cohere.com/changelog)) | Generally available |
| [GLM-5.3](https://theagenticweekly.com/models/glm-5-3) | Z.ai | GLM-5.3 License (MIT-style plus a MaaS condition) ([source](https://huggingface.co/zai-org/GLM-5.3/blob/main/LICENSE)) | 1.05M tokens ([source](https://docs.z.ai/guides/llm/glm-5.3)) | $1.4 in / $4.4 out per 1M tokens ([source](https://docs.z.ai/guides/overview/pricing)) | Aug 14, 2026 ([source](https://z.ai/blog/glm-5.3)) | Generally available |
| [Qwen3.8-27B](https://theagenticweekly.com/models/qwen3-8-27b) | Alibaba | Apache-2.0 ([source](https://huggingface.co/Qwen/Qwen3.8-27B)) | 262K tokens ([source](https://huggingface.co/Qwen/Qwen3.8-27B)) | Not published | Aug 14, 2026 ([source](https://github.com/QwenLM/Qwen3.8)) | Generally available |
| [Qwen3.8-2.4T-A95B](https://theagenticweekly.com/models/qwen3-8-2-4t-a95b) | Alibaba | Qwen3.8-Max License (custom) ([source](https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B/blob/main/LICENSE)) | 262K tokens ([source](https://huggingface.co/Qwen/Qwen3.8-2.4T-A95B)) | Not published | Aug 12, 2026 ([source](https://github.com/QwenLM/Qwen3.8)) | Generally available |
| [Kimi K3](https://theagenticweekly.com/models/kimi-k3) | Moonshot AI | Kimi K3 License (custom) ([source](https://huggingface.co/moonshotai/Kimi-K3)) | 1M tokens ([source](https://www.kimi.com/blog/kimi-k3)) | $3 in / $15 out per 1M tokens ([source](https://platform.kimi.ai/docs/pricing/chat)) | Jul 16, 2026 ([source](https://www.kimi.com/blog/kimi-k3)) | Generally available |
| [North Mini Code](https://theagenticweekly.com/models/north-mini-code) | Cohere | Apache 2.0 ([source](https://docs.cohere.com/changelog)) | 256K tokens ([source](https://docs.cohere.com/changelog)) | Not published | Jun 9, 2026 ([source](https://docs.cohere.com/changelog)) | Generally available |
| [Command A+](https://theagenticweekly.com/models/command-a-plus) | Cohere | Apache 2.0 ([source](https://docs.cohere.com/changelog)) | 128K tokens ([source](https://docs.cohere.com/changelog)) | Not published | May 20, 2026 ([source](https://docs.cohere.com/changelog)) | Generally available |
| [Mistral Medium 3.5](https://theagenticweekly.com/models/mistral-medium-3-5) | Mistral AI | Modified MIT ([source](https://docs.mistral.ai/getting-started/changelog)) | 256K tokens ([source](https://docs.mistral.ai/models/mistral-medium-3-5-26-04)) | $1.5 in / $7.5 out per 1M tokens ([source](https://docs.mistral.ai/models/mistral-medium-3-5-26-04)) | Apr 28, 2026 ([source](https://docs.mistral.ai/getting-started/changelog)) | Generally available |
| [Qwen3.6-27B](https://theagenticweekly.com/models/qwen3-6-27b) | Alibaba | Apache-2.0 ([source](https://huggingface.co/Qwen/Qwen3.6-27B)) | 262K tokens ([source](https://huggingface.co/Qwen/Qwen3.6-27B)) | Not published | Apr 22, 2026 ([source](https://github.com/QwenLM/Qwen3.8)) | Generally available |
| [GLM-5.1](https://theagenticweekly.com/models/glm-5-1) | Z.ai | MIT ([source](https://huggingface.co/zai-org/GLM-5.1)) | 200K tokens ([source](https://docs.z.ai/guides/llm/glm-5.1)) | $1.4 in / $4.4 out per 1M tokens ([source](https://docs.z.ai/guides/overview/pricing)) | Apr 7, 2026 ([source](https://docs.z.ai/release-notes/new-released)) | Generally available |
| [Gemma 4](https://theagenticweekly.com/models/gemma-4) | Google | Apache 2.0 ([source](https://blog.google/innovation-and-ai/technology/developers-tools/gemma-4/)) | 256K tokens ([source](https://ai.google.dev/gemma/docs/core)) | Not published | Mar 31, 2026 ([source](https://ai.google.dev/gemma/docs/releases)) | Generally available |
| [Mistral Small 4](https://theagenticweekly.com/models/mistral-small-4) | Mistral AI | Apache 2.0 ([source](https://docs.mistral.ai/models/mistral-small-4-0-26-03)) | 256K tokens ([source](https://docs.mistral.ai/models/mistral-small-4-0-26-03)) | $0.15 in / $0.6 out per 1M tokens ([source](https://docs.mistral.ai/models/mistral-small-4-0-26-03)) | Mar 16, 2026 ([source](https://docs.mistral.ai/getting-started/changelog)) | Generally available |
| [GLM-5](https://theagenticweekly.com/models/glm-5) | Z.ai | MIT ([source](https://huggingface.co/zai-org/GLM-5)) | 200K tokens ([source](https://docs.z.ai/guides/llm/glm-5)) | $1 in / $3.2 out per 1M tokens ([source](https://docs.z.ai/guides/overview/pricing)) | Feb 12, 2026 ([source](https://docs.z.ai/release-notes/new-released)) | Generally available |
| [Ministral 3 14B](https://theagenticweekly.com/models/ministral-3-14b) | Mistral AI | Apache 2.0 ([source](https://docs.mistral.ai/models/ministral-3-14b-25-12)) | 256K tokens ([source](https://docs.mistral.ai/models/ministral-3-14b-25-12)) | $0.2 in / $0.2 out per 1M tokens ([source](https://docs.mistral.ai/models/ministral-3-14b-25-12)) | Dec 2, 2025 ([source](https://docs.mistral.ai/getting-started/changelog)) | Generally available |

## FAQ

**What is the best open-weight LLM?** It depends on your hardware and use. Newest frontier open-weight model: Muse Spark Open. Largest context window: GLM-5.3. Lowest API price, if you do not self-host: MiniMax M3.1 Flash. Sized for your own hardware: Nemotron 3.5 Bolt E4B (4GB).

**Are open-weight models open source?** Not always. Some use permissive licences such as Apache 2.0 or MIT; others add use restrictions. Each model's licence is in the table, linked to the vendor's page.

**Which open-weight models can run on my own machine?** Models we file under "runs on your own hardware": Nemotron 3.5 Bolt E4B (4GB), Qwen3.8-27B, North Mini Code, Qwen3.6-27B, Gemma 4. Check the parameter count and the vendor's hardware notes.

**Did you run these models?** No. This page is Reported from the vendors' model cards, licences and pricing pages.

## How to cite

Agentic Weekly, "Best open-weight LLMs: how to choose", updated 2026-10-29, https://theagenticweekly.com/best/open-weight-llms