Open-source LLMs stopped being the budget option in 2026. Kimi K3 sits level with Claude Opus 4.8 on the Artificial Analysis Intelligence Index (its hosted API is live; the weights themselves are expected by July 27), GLM-5.2 held the top open-model spot before it, and the field behind them is deep enough that the hard part is choosing.
This ranking covers the nine best open-weight models right now — on license, context window, hardware reality, and the per-token price you actually pay. Every one of them is available through LLM Gateway with one key, at each provider's published rate, so you can A/B any two of them by changing one word in a request.
1. Kimi K3 — the open frontier
Moonshot AI · 2.8T params · 1M context · $3.00 / $15.00 per M
The largest open-weight model ever announced — with one caveat: the weights are not downloadable yet. Moonshot expects to release them by July 27, 2026, and the license is still unannounced; the hosted API has been live since July 16. Ranks 4th of 189 models on the Artificial Analysis Intelligence Index — tied with Claude Opus 4.8 and GPT-5.5 — and took first place in Arena's blind Frontend Code testing. Always-on reasoning, vision, tools, and output configurable up to 1M tokens. The open model to beat, priced accordingly. .
How to choose
Maximum capability: Kimi K3, with GLM-5.2 as the value alternative
Production volume: DeepSeek V4 Pro or gpt-oss-120b — pennies per million
Multimodal agents: MiniMax M3, or Qwen3.6-35B-A3B at the small end
Self-hosting: match the model to your hardware — Qwen3.6-35B-A3B for a workstation, gpt-oss-120b for a single 80 GB GPU, GLM-5.2 for a full node
Coding agents: K3 for the hard problems, GLM-5.2 or DeepSeek V4 Pro for the loop — : flat $29/$79/$179 a month for coding agents, with roughly 3× your subscription price in model usage at provider rates. On DevPass, every model on this list except Kimi K3 is standard-tier with no weekly cap; K3 crosses the premium price threshold and draws from a weekly premium allowance.
Frequently Asked Questions
What is the best open-source LLM in 2026?
Kimi K3, by benchmark standing: 4th of 189 models on the Artificial Analysis Intelligence Index, level with Claude Opus 4.8 — though its weights are expected by July 27 and it is API-only until then. If price matters, GLM-5.2 delivers most of that capability with output at under a third of K3's price.
What is the best small open-source LLM?
Qwen3.6-35B-A3B — Apache 2.0, multimodal, 3B active parameters, and it runs on a 24 GB consumer GPU while holding its own on agentic coding benchmarks.
Are these models open source or open weight?
Open weight: the trained weights are downloadable and self-hostable under permissive licenses (mostly MIT or Apache 2.0), but training data and code are generally not released. GLM-5.2, DeepSeek V4 Pro, Qwen3.6-35B-A3B, and gpt-oss-120b use standard OSI licenses; Llama uses Meta's community license. Kimi K3 is the exception for now: its weights are expected by July 27, 2026 and are unlicensed until Moonshot announces terms — today it is usable only through hosted APIs.
What is the cheapest way to use open-source LLMs?
Through a gateway at provider list prices. LLM Gateway adds no per-token markup — pay-as-you-go credits carry a 5% platform fee at top-up, and DevPass turns heavy coding-agent usage into a flat monthly rate.
Getting started
— flat-rate open models in your coding agent from $29/mo- Deep dive on the leader: .↗ Original-Artikel auf dev.to lesenVollständiger Original-BerichtAusführliche Details, Code-Beispiele & Hersteller-Stellungnahme auf dev.to.
SOCIAL SHARE CARD GENERATOR