model rankings

Best Open-Source LLMs

The best open-source and open-weight LLMs in 2026 — DeepSeek V4.1-Flash, Kimi K3.1 Pro, GLM-5.3, Kimi K3, Mistral Large 4, Muse Spark Open, DeepSeek V4, MiniMax M3, GLM-5.2, Qwen 3.6, Llama 4, gpt-oss and Gemma 4 ranked by benchmark, licence, context window and cost, with self-hosting and hosted-inference picks for every use case.

Updated 6 November 2026

Quick answer: The best open-source LLM you can run today is DeepSeek V4.1-Flash (10 September 2026), because it pairs an MIT licence and a 1M-token context with 39 on the independent Artificial Analysis Intelligence Index (v4.3.2) for its September launch build, above DeepSeek’s own 1.6T-parameter V4-Pro, at $0.15/$0.60 per million tokens off-peak or free to self-host. The most capable open model is Moonshot’s Kimi K3.1 Pro, whose weights shipped on 4 November: Artificial Analysis scored it 55 on the same index version at its 23 October launch, ten points above GLM-5.3, under an MIT-based licence with conditions for the largest businesses. It is a 2.8T-parameter, 1.56 TB self-host and cost about eight times as much per completed v4.3.2 index task, which keeps V4.1-Flash, at 552B parameters, the best all-round pick for most teams. Z.ai’s GLM-5.3 (custom licence, text-only) and Moonshot’s earlier Kimi K3 follow. Mistral’s Mistral Large 4 is the strongest open model from outside China: when its weights were released on 29 October, Artificial Analysis scored it 26 on its newer v5.0 index, a point above Meta’s Muse Spark Open, the strongest from a US lab. DeepSeek V4 still posts the highest open SWE-bench Verified score (80.6%, vendor-reported), though DeepSeek’s API now answers V4 requests with V4.1-Flash. MiniMax M3 is the closest practical alternative for high-throughput, long-context work (80.5% Verified, 1M context). GLM-5.3 is also the strongest pick for long-horizon agents, with the MIT-licensed GLM-5.2 as the plain-licence alternative; Alibaba’s Apache 2.0 Qwen3.8-27B is the strongest small open model; Qwen 3.6 and Llama 4 have the widest fine-tuning and tooling support; and Gemma 4 is the best model that runs on a single laptop, and Meta’s new Apache 2.0 Muse Glimmer is the pick for local agents on a 24GB GPU. New since mid-July: Thinking Machines’ Apache 2.0 Inkling and Inkling-Small are the first open models to reason natively over text, images and audio — the pick if your self-hosted workload is multimodal. The one caveat: “open source” is mostly a misnomer here — nearly all of these ship as open weights under custom licences, not fully open-source projects, so read the licence before you build a business on one.

This is an opinionated ranking of models you can download, inspect, fine-tune and run on your own hardware. It is not a scrape of any single leaderboard. We rank by consensus — an aggregate of independent testing (chiefly Artificial Analysis), vendor benchmarks, licences and cost — and we cite the evidence behind every placement. Where a figure isn’t verifiable from a primary source, we mark it “data not available” rather than invent one. For the full field including proprietary frontier models, see our best AI models ranking; for the coding-specific view, see best AI for coding.

Open source vs open weight: the distinction that matters

Most models marketed as “open source” are not open source in the strict sense. Understanding the difference decides whether you can legally ship a product on one.

Open weight means the trained model weights are published for download, but the training data, training code and full pipeline may not be. Llama, Qwen, DeepSeek, GLM, Kimi and Gemma all sit here. You can run and fine-tune the model, but you cannot fully reproduce it.

Fully open source means weights, training code, training data and an OSI-approved licence are all public. Very few frontier-adjacent models qualify; Allen AI’s OLMo line is the usual reference point for genuinely open training.

The licences vary more than the marketing suggests. MIT (used by DeepSeek V4 and GLM-5.2) and Apache 2.0 (used by gpt-oss, the Mistral 3 family and, since April 2026, Gemma 4) are the cleanest for commercial use. Meta’s Llama Community Licence permits commercial use only up to 700 million monthly active users, above which an explicit Meta agreement is required. The practical rule: if you are building a commercial product, prefer an MIT- or Apache-licensed model, and read any custom licence in full before committing.

For most teams the honest reason to pick an open model isn’t ideology — it’s control. Open weights buy you self-hosting, data residency, offline operation, unlimited fine-tuning, no per-token bill and no vendor able to deprecate the model out from under you. Those are the axes this ranking weighs.

The ranking: best open-source LLMs

These are the strongest models you can download and run, ranked by overall capability and consensus standing, weighted toward repository-scale coding and agentic work (where the most comparable cross-model data exists), then reasoning, context and licence quality. For how they compare with closed models, see our master best AI models board; here we rank the open-weight field on its own terms, including the largest open-weight models, Kimi K3.1 Pro, GLM-5.3, Kimi K3 and Mistral Large 4.

ᵛ Vendor-reported. Muse Glimmer’s figures come from Meta’s own model card with no independent evaluation yet, which is why a 30B model that beats Qwen 3.6 27B on most agentic benchmarks sits at #16 rather than higher — we treat vendor numbers as a ceiling.

Reading the table. SWE-bench Verified is the older, Python-only coding benchmark still quoted at most launches (llm-stats); we show it where a verified figure exists and mark “data not available” where the model’s own page is still in build rather than invent a number. Context is the maximum published window. Hosted prices are the cheapest first-party or major-host rate we can verify in USD per million tokens; every model here can also be self-hosted, where the only cost is your own compute. Licences are the weight licence, not the code licence — see the licence table below.

One number worth sitting with: DeepSeek V4, MiniMax M3 and Alibaba’s closed Qwen3.7 Max all land within 0.2 points of each other (80.6%, 80.5%, 80.4%) on SWE-bench Verified, and all three are within a fraction of a point of Google’s proprietary Gemini 3.1 Pro. The open frontier has effectively caught last year’s closed leaders, at roughly a tenth of the price.

Segmented verdicts

The single-number ranking hides the fact that “best” depends on the job. Here are the decisive picks.

Best all-round open model: DeepSeek V4.1-Flash

DeepSeek V4.1-Flash is the model to reach for first. Artificial Analysis scored its September launch build 39 on its Intelligence Index v4.3.2 at max effort, and its launch post put it above DeepSeek V4-Pro, a model three times its size. The MIT licence makes it self-hostable outright — no usage caps, no competitive-use clause, no MAU ceiling — and at 552B parameters (8B active per token for input, 16B for output) it is a third the size of V4-Pro, with a 1M-token context and native image input. Hosted on DeepSeek’s API it costs $0.15/$0.60 per million tokens off-peak, double in peak hours, and Artificial Analysis measured $0.27 per completed index task (DeepSeek). Two caveats: DeepSeek published no SWE-bench figure, so its coding case rests on its own Terminal-Bench 2.1 (90.6) and DeepSWE v1.1 (74.2) results, and Artificial Analysis measured it as verbose. DeepSeek V4 keeps the higher vendor-reported SWE-bench Verified score (80.6%) and its weights remain downloadable, but DeepSeek’s API now answers V4 requests with V4.1-Flash.

Highest raw capability: Kimi K3.1 Pro

Kimi K3.1 Pro is the most capable open-weight model on the independent evidence. Artificial Analysis scored it 55 on its Intelligence Index v4.3.2 at its 23 October API launch, ten points above GLM-5.3 on that version; its current score is on the best AI models board. Moonshot published the weights on Hugging Face on 4 November 2026: 2.8 trillion parameters with 104 billion active, image input and a 1,048,576-token context, shipped in MXFP4 at about 1.56 TB across 96 shards. The licence is MIT-based but not MIT: a model-as-a-service business with more than $20 million in revenue over 12 months needs a separate agreement with Moonshot before commercial use, and a product above 100 million monthly active users or $20 million in monthly revenue must display “Kimi K3.1-Pro” in its interface; internal use is exempt. The caveats: self-hosting it needs a multi-GPU cluster, Moonshot publishes no SWE-bench figure, and Artificial Analysis measured it as slow and verbose, at $2.11 per completed v4.3.2 index task against V4.1-Flash’s $0.27. Hosted, it costs $2/$10 per million tokens on the Kimi API.

GLM-5.3 is next on the independent evidence: Artificial Analysis scored it 60 on its Intelligence Index at its 14 August launch, level with Kimi K3 on Artificial Analysis’s August run, and it is the smaller and cheaper of the two. Z.ai held the weights back at launch over the model’s exploit-finding ability and published them on Hugging Face on 28 August 2026 under a custom GLM-5.3 Licence: MIT-style permissions, plus a Z.ai security review before commercial use by model-as-a-service businesses with more than $10 billion in revenue. At 753B parameters it is far easier to self-host than either Kimi model’s 2.8 trillion, and it costs $1.40/$4.40 hosted. The caveats: it is text-only, it publishes no SWE-bench figure, and it costs several times more per token than V4.1-Flash.

Kimi K3, open-weight since 26 July, scored 57.1 on Artificial Analysis’s Intelligence Index v4.1 at its 16 July launch and has the same 2.8 trillion parameters (~1.56TB) as Kimi K3.1 Pro, hosted at $3/$15 per million tokens; Kimi K3.1 Pro supersedes it for most uses at a lower token price. If you want the strongest open model you can practically self-host cheaply, DeepSeek V4.1-Flash is the better answer.

Strongest open model from outside China: Mistral Large 4

Mistral Large 4 is Mistral’s 1.05-trillion-parameter mixture-of-experts model (52B active), with image input and a 1M-token context. It opened as an API preview on 6 October 2026, and on 29 October Mistral published updated weights on Hugging Face and cut the API price to $0.70/$2.10 per million tokens (Mistral). Artificial Analysis scored the release 26 on its Intelligence Index v5.0 when the weights came out on 29 October, and called it the highest-scoring open model developed outside China; on that version at the time, DeepSeek V4.1-Flash and GLM-5.3 scored 32 and DeepSeek V4-Pro 23. The preview scored 38 on v4.3.2 at its 6 October launch. Mistral’s own figures include 41% on Terminal-Bench 4.0 and 48% on the Artificial Analysis Coding Agent Index, and it publishes no SWE-bench score. The licence is a modified MIT licence: free to use, fine-tune and build products on, but a business offering the model as a hosted service needs a separate commercial licence from Mistral.

Strongest US open model: Muse Spark Open

Muse Spark Open is the open-weight model Meta promised when it took Muse Spark proprietary. Released on 15 October 2026, it is an 816-billion-parameter mixture-of-experts model (42B active) with a 1M-token context, on Hugging Face and through API hosts including OpenRouter, Groq, Fireworks AI and Together AI (Meta). Artificial Analysis scored it 39 on its Intelligence Index v4.3 at xhigh at its launch, level with DeepSeek V4.1-Flash, and when Artificial Analysis moved to v5.0 on 27 October it scored 25 there, a point under Mistral Large 4 and, on Artificial Analysis’s reading at the time, the highest-scoring US-developed open model. Thinking Machines’ Inkling 2 Small scored 31 on v5.0.1 at its 5 November launch, but its weights were closed at launch. Meta published no SWE-bench figure. The catch is the licence: Meta calls the Muse Model License’s commercial terms “limited”, and DeepLearning.AI reports that companies with more than $20 million in revenue selling access to the model need a separate agreement with Meta.

Best for high throughput and long context: MiniMax M3

MiniMax M3 sits a hair behind DeepSeek V4 on SWE-bench Verified (80.5%) with the same 1M-token context, and its mixture-of-experts design is tuned for cheap, high-volume serving. Hosted at $0.30/$1.20 per million tokens, it is the pick when you are pushing large batches of long inputs through and want open weights you can also pull in-house later.

Best for long-horizon agents: GLM-5.3

GLM-5.3 is the strongest open option for multi-step, tool-using agent runs. Z.ai built it by scaling post-training on long agentic trajectories on top of GLM-5.2’s base model, and its own table moves Terminal-Bench 3.0 from 4.6 to 28.3 and DeepSWE v1.1 from 46.2 to 66.9 (Z.ai), vendor figures backed by Artificial Analysis’s independent index, which scored it 60 at its August launch. Its licence adds a security review only for model-as-a-service businesses above $10 billion in revenue. If you need a plain MIT licence, GLM-5.2 is the same base model with a 1M-token context and lighter post-training.

Z.ai also released GLM-5.3-Flash under MIT on 26 August: a separately trained 320B MoE (18B active) taking text, images and video at a 1M context, and the smaller, multimodal option in the line.

Best multimodal open model: Inkling

Inkling and Inkling-Small (Thinking Machines Lab, July 2026, Apache 2.0) are the first open-weight models that reason natively over text, images and audio in one model — the Kimi models stop at vision, and Z.ai’s MIT-licensed GLM-5.3-Flash (26 August, 320B/18B active, images and video at a 1M context) is the newest multimodal alternative, though it takes no audio. The surprise is that the 276B Small beats its 975B parent on SWE-bench Verified (80.2% vs 77.6%, vendor), so it is the one to deploy: a quarter of the download, 12B active parameters, and a 1M context. The honest caveat: Artificial Analysis scored Inkling 42 at its July launch and GLM-5.3 60 at its August launch, both before v4.3 came out on 7 September — you choose it for the modalities and the clean licence, not for peak text intelligence. It is also built to be fine-tuned, via the lab’s Tinker service. Its successor, Inkling 2 Small (5 November 2026), takes text and images but not audio, and launched with closed weights.

Best for fine-tuning and ecosystem support: Qwen 3.6 and Llama 4

For customisation, the deciding factor is how much tooling, documentation and community fine-tuning support already exists. Qwen 3.6 (Apache 2.0) and Llama 4 (Llama Community Licence) win here: both have the broadest coverage in training frameworks, quantisation formats and adapter libraries, so the path from base model to a task-specific fine-tune is the best-trodden. Alibaba’s newest open Qwen is Qwen3.8-27B (August 2026, Apache 2.0), which takes text, image and video input; Artificial Analysis scored it 34 on its Intelligence Index v4.3.2 after its August release, then the highest of any open model its size; Alibaba also published a text-only checkpoint of the 2.4T Qwen3.8-Max under a more restrictive licence.

Best for the largest context: Llama 4

Llama 4 Scout reaches a 10M-token context window — the single largest of any model on this list, open or closed. For workloads that genuinely need to hold an entire codebase, book series or archive in context at once, it is the open pick, subject to the Llama Community Licence’s 700M-MAU commercial ceiling.

Best on a laptop or at the edge: Gemma 4

Gemma 4 is the standout for running locally on modest hardware. Its smaller variants are widely rated the most practical model to run on a single laptop or consumer GPU, making it the default for offline, private, on-device use. Gemma 4 also dropped the custom Gemma Terms of Use that constrained earlier generations: since the 2 April 2026 release the weights ship under Apache 2.0, so commercial use, fine-tuning and redistribution carry no royalties, no MAU ceiling and no bespoke agreement with Google. For the hardware side — which variant fits which card — see our best local LLMs guide.

Best open reasoning model: DeepSeek R1 / R2

For explicit chain-of-thought reasoning in open weights, DeepSeek R1 / R2 is the reference — an MIT-licensed reasoning line you can self-host for maths, science and multi-step logic where you want the model to show its working.

Best for a clean commercial licence: gpt-oss and the Mistral 3 family

If licence risk is your first concern, gpt-oss 120b (Apache 2.0) and Mistral Large / Medium 3.5 (Apache 2.0) are the safest picks — permissive, OSI-approved, no MAU ceiling. Mistral additionally suits teams that need EU data residency and a European vendor relationship. Its bigger Mistral Large 4 is not on Apache 2.0: its licence adds a commercial-licence condition for hosted-service businesses.

Licence comparison: what you can legally ship

For a commercial product, the licence matters as much as the benchmark. This table summarises the weight licences, not the code repositories.

Bottom line: MIT and Apache 2.0 models (DeepSeek, GLM, gpt-oss, Qwen, Mistral, MiniMax, Gemma 4 and now Meta’s Muse Glimmer) carry the least legal friction. Llama’s licence is fine for the overwhelming majority of products but has a scale trigger worth knowing about — and note that Meta now ships under three licences: Llama’s MAU ceiling, Apache 2.0 for Muse Glimmer, and the Muse Model License for Muse Spark Open. Any vendor-specific “open weights” licence — Kimi’s, GLM-5.3’s, the Qwen3.8 2.4T checkpoint’s, Mistral Large 4’s and Muse Spark Open’s included — should be read in full before you depend on it commercially; GLM-5.3’s, Qwen3.8’s and Kimi K3.1 Pro’s add conditions only for the largest companies, Mistral’s only for businesses selling the model as a hosted service, and Meta’s for companies above $20 million in revenue selling access to it.

How to run open-source LLMs

You do not need your own GPUs to use these models. There are two paths: self-host the weights, or call an open model through a hosted-inference provider that has already done the serving for you. The table below is built to be affiliate-ready; we take no commission and links are plain.

Practical guidance. For a laptop or a quick local trial, start with Ollama or LM Studio and a small model like Gemma 4 or a quantised DeepSeek variant. For production self-hosting, vLLM on your own GPUs gives the best throughput and an OpenAI-compatible endpoint. If you want open-model economics without running hardware, a hosted provider (Together, Fireworks, OpenRouter or Groq) lets you call DeepSeek V4 or MiniMax M3 by API. One serving note that materially affects quality: prefer bf16 over fp8 activation quantisation where you can, because aggressive quantisation degrades output on these models. Our companion best local LLMs page covers the hardware-and-deployment side in full — VRAM requirements by model, quantisation formats and what actually runs on consumer machines from an 8GB laptop up to a 128GB unified-memory desktop.

What changed this month

The freshness signal — the open-weight movements that reshaped the board through mid-2026.

Thinking Machines released Inkling 2 Small with closed weights (5 November) — not ranked here until they ship. Inkling 2 Small, a 594B-parameter mixture-of-experts model with 17B active and image input, launched on Thinking Machines’ Tinker service only and scores 31 on the Artificial Analysis Intelligence Index v5.0.1. Thinking Machines says it intends to release the weights under Apache 2.0 after external safety testing and red-teaming (Thinking Machines); it joins this ranking when the weights and licence are published.

Moonshot published Kimi K3.1 Pro’s weights (4 November) — the new capability leader, at #2. Kimi K3.1 Pro is on Hugging Face under the MIT-based Kimi K3.1 Pro License: 2.8T total and 104B active parameters, about 1.56 TB in MXFP4. At 55 on the Artificial Analysis index at launch (46 on v5.0) it passes GLM-5.3 as the most capable open model. Its size, cost per task and licence conditions keep DeepSeek V4.1-Flash at #1; everything from GLM-5.3 down moves one place.

DeepSeek shipped V4.1-Flash (10 September) — the new #1. DeepSeek V4.1-Flash replaced V4-Flash on DeepSeek’s API under the MIT licence: a 552B-parameter model with image input and a 1M-token context at $0.15/$0.60 off-peak. Artificial Analysis scored it 39, above V4-Pro, and DeepSeek is phasing V4-Pro out, routing its API requests to V4.1-Flash since 14 September. It takes #1 as the best all-round open model, and DeepSeek V4 moves down.

GLM-5.3’s weights shipped (28 August) — the new capability leader. Z.ai published GLM-5.3 on Hugging Face under a custom GLM-5.3 Licence after a two-week safety hold. At 45 on the independent index it passes Kimi K3 (44) as the most capable open model and enters at #2; Kimi K3 moves to #3.

Qwen3.8’s open weights shipped (August). Alibaba published Qwen3.8-27B under Apache 2.0, now the strongest small open model at 34 on the index, and a text-only checkpoint of the 2.4T Qwen3.8-Max under its own licence. Qwen3.8-27B enters at #7.

Meta came back to open source — under Apache 2.0 (10 August). Meta released Muse Glimmer, a 30B dense multimodal model distilled from the proprietary Muse Spark 1.2 and built for local agentic work, under a plain Apache 2.0 licence — more permissive than any Llama model ever carried, with no MAU ceiling. It runs on a 24GB consumer GPU at 4-bit (~17GB) and beats Qwen 3.6 27B and Gemma 4 31B on most agentic benchmarks on Meta’s numbers, while trailing Qwen on computer use and terminal work. Two things to hold: every figure is vendor-reported, and at 30B it is not competing with the top of this board. The bigger prize was the promise attached to it, open weights for Muse Spark itself, which Meta delivered on 15 October with Muse Spark Open.

The open frontier caught the closed leaders. DeepSeek V4 (MIT), MiniMax M3 and Alibaba’s Qwen3.7 Max now all sit within ~0.2 points of Gemini 3.1 Pro on SWE-bench Verified, at roughly a tenth of the cost. The gap between open and proprietary that used to be 40–60 points wide is now a handful.

Kimi K3’s open weights shipped (26 July). Moonshot released the weights for Kimi K3 — the largest open model ever at 2.8T parameters and, on the independent intelligence index, the new raw-capability leader among open models. Its size makes it a serious-hardware deploy, so DeepSeek V4 stays the most practical open default, but the open capability ceiling moved up overnight.

Qwen’s top tier went closed — then started to reopen. Alibaba took Qwen3.7 Max to API-only, so the strongest Qwen was no longer downloadable and the open Qwen you can self-host became the 3.6 line. That reversed with Qwen3.8-Max (3 August): in August Alibaba published a text-only checkpoint of the 2.4T model under its own licence and Qwen3.8-27B under Apache 2.0 — a reminder that “open” status can change release to release, and worth checking before you standardise on a model.

Licences kept converging on the permissive end. The clear trend across 2025–2026 has been makers moving to MIT and Apache 2.0: DeepSeek to MIT with V3/R1, Qwen to Apache 2.0 with the Qwen3 generation, and new entrants such as gpt-oss and the Mistral 3 family launching directly under Apache 2.0. Gemma 4 joined them in April 2026, moving from Google’s custom Gemma Terms of Use to Apache 2.0. For commercial builders, open-weight licence risk is lower than it was a year ago — Llama’s MAU ceiling and the vendor-specific Kimi licences are now the main exceptions to read carefully.

Self-hosting got easier. Tooling like Ollama, LM Studio and vLLM has matured to the point where running a capable open model locally or in production is a matter of hours, not weeks — lowering the practical barrier that used to keep teams on closed APIs.

How we rank

Our methodology is deliberately plural, because no single source is trustworthy on its own.

We start with benchmarks, read critically: we quote SWE-bench Verified where a verified figure exists, label it as the older Python-only measure it is, and mark “data not available” rather than guess where a model’s page is still in build. We weight memorisation-resistant and unsaturated evals over ones the top models have already saturated.

We layer in licence quality, because an open model you cannot legally ship is not useful, and independent testing, chiefly the Artificial Analysis Intelligence Index, which we weight above any lab’s own numbers. We run no benchmarks of our own. And we weight operational reality: context window, hosted price, and how much tooling exists to run and fine-tune the model.

Every figure on this page is either cited to a primary source or linked to our own model pages, where the detail lives. Where a number can’t be verified, we write “data not available”. This ranking is updated monthly; the date at the top is the last full pass. The honest limitation: several open-weight models here have no standardised, independently verified benchmark figure yet — we mark those cells rather than fill them with vendor claims.

Frequently asked questions

What is the best open-source LLM in 2026?

The best open-source LLM you can run today is DeepSeek V4.1-Flash — MIT-licensed, 1M-token context, and 39 on the independent Artificial Analysis Intelligence Index (v4.3.2) for its September launch build, above DeepSeek’s larger V4-Pro. It is free to self-host and cheap to run hosted ($0.15/$0.60 per million tokens off-peak). Kimi K3.1 Pro, open-weight since 4 November, is the most capable open model (55 on the same index at launch), but it needs serious hardware to self-host and its MIT-based licence adds conditions for the largest businesses; GLM-5.3 (text-only, custom licence) follows. MiniMax M3 is the closest alternative for high-throughput, long-context work.

Is there a difference between open source and open weight?

Yes, and it matters commercially. Open weight means the model’s trained weights are downloadable, but its training data and pipeline may not be — this covers Llama, Qwen, DeepSeek, GLM, Kimi and Gemma. Fully open source means weights, training code, training data and an OSI-approved licence are all public, which very few frontier-adjacent models are (Allen AI’s OLMo is the usual example). Most models marketed as “open source” are technically open weight.

Are open-source LLMs as good as GPT or Claude now?

For most routine work, close. DeepSeek V4, MiniMax M3 and Alibaba’s Qwen line sit within ~0.2 points of Google’s proprietary Gemini 3.1 Pro on SWE-bench Verified, at roughly a tenth of the cost. Proprietary flagships like Claude Opus 5 still lead on the hardest tasks and on long-horizon agentic reliability, but paired with a strong harness an open model handles the bulk of everyday tasks.

Which open-source LLM has the best licence for commercial use?

MIT and Apache 2.0 models carry the least legal friction. DeepSeek V4.1-Flash, DeepSeek V4 and GLM-5.2 are MIT; gpt-oss, Qwen 3.6, Gemma 4, Inkling and Mistral Large / Medium 3.5 are Apache 2.0 — all permit unrestricted commercial use. Llama 4’s Community Licence is the main exception, permitting commercial use only up to 700 million monthly active users.

Can I run an open-source LLM on my own computer?

Yes. For a laptop or single GPU, install Ollama or LM Studio and pull a smaller model such as Gemma 4 or a quantised DeepSeek variant. For production self-hosting on your own GPUs, vLLM gives high throughput and an OpenAI-compatible API. If you want open-model economics without owning hardware, hosted providers such as Together AI, Fireworks AI, OpenRouter and Groq serve open models by API.

What is the best open-source LLM for coding?

DeepSeek V4.1-Flash is the default open coding model: on DeepSeek’s own table it scores 74.2% on DeepSWE v1.1 and 90.6 on Terminal-Bench 2.1, though it has no SWE-bench figure. DeepSeek V4 holds the highest open SWE-bench Verified score (80.6%, vendor-reported), with Kimi K2.6 strong for agentic coding workflows and Qwen 3.6 well supported for fine-tuning on your own codebase. For the complete picture including proprietary tools and agents, see our best AI for coding guide.

What is the best open-source LLM for agents?

GLM-5.3 is the strongest open pick for long-horizon, tool-using agents you can practically self-host; its licence adds a security review only for model-as-a-service businesses above $10 billion in revenue. For a plain MIT licence, GLM-5.2 is the same base model with lighter post-training, and Kimi K2.6 is an alternative for agentic and coding tasks. Trial them on your specific tool-calling workload before committing.

Which open-source LLM has the largest context window?

Llama 4 Scout has the largest published context window of any model on this list at up to 10 million tokens. DeepSeek V4.1-Flash, DeepSeek V4 and MiniMax M3 all offer 1 million tokens, which is enough for whole-repository or long-document work for most teams.