Open-weight models in October 2026 are led by Chinese labs, with Europe and Meta catching up. DeepSeek V4.1 Flash (MIT licence) is now DeepSeek's top API model and the most practical open-weight default. Kimi K3 is the strongest open-weight model for hard reasoning, at a higher price. GLM-5.3 and Qwen 3.8 Max are strong coders with 1M-token windows. Mistral Large 4, a ~1 trillion-parameter European model, is in preview with weights promised for late October, and Meta's Muse Glimmer 30B is the best small model under Apache 2.0.
The safest this comparison default, the cheaper option worth trying first, and the specialist pick — before you read the detail below.
DeepSeek V4.1 Flash is the strongest answer here for this comparison — pick it when quality of output matters more than the $0.30/1M/1M input you pay for it.
Claude Sonnet 5.5 is the cheaper way in for this comparison, at $2.00/1M/1M input against DeepSeek V4.1 Flash's $0.30/1M/1M.
Kimi K3 is the fastest of these for this comparison — worth it when latency is what the reader notices, not the last few points of reasoning depth.
DeepSeek V4.1 Flash is the most practical open-weight model: MIT-licensed, multimodal and cheap to host.
Kimi K3, GLM-5.3 and Qwen 3.8 Max close most of the gap to closed models on coding, each with a 1M-token window.
The best closed models — Claude Opus 5.5, GPT-6 Astra — still lead on long agent runs and the hardest tasks.
Choose DeepSeek V4.1 Flash for a general open-weight default you can host anywhere.
Choose Kimi K3 when reasoning depth matters more than cost — it is open-weight but priced like a premium model on hosted APIs.
Choose Mistral Medium 3.5 (or Large 4 once released) for European hosting, and Muse Glimmer 30B for local use on one GPU.
Switch the scoring lens to see whether the this comparison answer changes when cost, speed, or long-document depth leads the decision.
Moonshot / Premium / Aug 6, 2026
Closest Chinese challenger to the frontier — #4 overall on intelligence.
Ranks models by the broadest mix of coding, writing, research, and long-context usefulness.
You need fast responses or predictable output costs — always-on thinking burns tokens.
Where the this comparison recommendation shifts once you weigh price or latency differently.
DeepSeek's open, cheap new default — V4-Pro now routes to it.
Closest Chinese challenger to the frontier — #4 overall on intelligence.
Same price as GLM-5.2, far stronger on agents and security.
Best Chinese flagship — beats GPT-5.6 Sol on coding, #2 globally for vision.
Best self-hostable multimodal model — European, dense, MIT-licensed.
Apache 2.0 agent model that runs on a 24GB GPU.
List prices and published scores — the numbers this page's pick is built from.
| Model | Input | Output | Est. month | Context | Speed | Coding | Writing | Research |
|---|---|---|---|---|---|---|---|---|
| DeepSeek V4.1 FlashDeepSeek | $0.30/1M | $1.20/1M | $5.40 | 1.0M tokens | Fast | 89 | 83 | 83 |
| Kimi K3Moonshot | $3.00/1M | $15.00/1M | $60 | 1M tokens | Deliberate | 96 | 90 | 93 |
| GLM-5.3Z.ai | $1.40/1M | $4.40/1M | $23 | 1M tokens | Balanced | 91 | 79 | 82 |
| Qwen 3.8 MaxAlibaba | $2.00/1M | $6.00/1M | $32 | 1M tokens | Balanced | 93 | 85 | 88 |
Scores out of 100 — how we evaluate models. “Est. month” is 10M in / 2M out at list price: a ceiling, no discounts.
Why each one is on the shortlist for this comparison, what it is genuinely good at, and where we would steer you away from it.
Ranked first here for this comparison: 89/100 on coding, with the widest margin of anything in this line-up.
DeepSeek's September 10, 2026 release, now its top API model: a natively multimodal, MIT-licensed model with 1M context that DeepSeek says beats its own V4-Pro on coding and agent work — V4-Pro traffic has been routed to it since September 14.
You need long single outputs (32K cap) or independently verified scores before you commit.
DeepSeek's current flagship in practice, since V4-Pro requests now run on it. Cheap and open, with coding claims that still need independent confirmation. A strong pick for self-hosting or budget pipelines with review.
Full pricing, benchmark table and release notes on the DeepSeek V4.1 Flash page.
The fastest model in this shortlist for this comparison. Pick it when turnaround is what your readers or users notice.
Moonshot's 2.8-trillion-parameter multimodal reasoning flagship with always-on thinking — the largest open-weight model ever released and the closest Chinese challenger to the Western frontier.
You need fast responses or predictable output costs — always-on thinking burns tokens.
The first Chinese model to genuinely crowd the Western frontier — #4 on aggregate intelligence ahead of Opus 4.8. The always-on thinking makes it slow and output-heavy, so cost per task runs above the sticker price. A serious Opus-class alternative if latency isn't critical.
Full pricing, benchmark table and release notes on the Kimi K3 page.
The alternative to check next for this comparison — 91/100 on coding.
Same price as GLM-5.2, far stronger on agents and security. Full GLM-5.3 review →
Also worth a look for this comparison, at 93/100 on the coding axis.
Best Chinese flagship — beats GPT-5.6 Sol on coding, #2 globally for vision. Full Qwen 3.8 Max review →
UseRightAI recommendations are based on practical decision factors people actually feel in day-to-day use.
Newsletter
We email when the this comparison pick changes, when one of these models moves on price, or when something new displaces the current leader.
No spam. Useful updates only. Affiliate disclosures always clearly labeled.
As of October 2026, DeepSeek V4.1 Flash is the best general open-weight model for most uses, and Kimi K3 the strongest for hard reasoning. Strictly, most of these are open-weight rather than fully open source: the weights are published, but the training data and code usually are not.
On many everyday tasks, nearly. The gap is largest on long autonomous agent runs and the hardest coding problems, where Claude Opus 5.5 and GPT-6 Astra still lead. Many open-weight benchmark claims are self-reported, so check independent results before relying on them.
Hosted: Groq, Together AI, Fireworks and similar providers, or the Vercel AI Gateway. Locally: Ollama or LM Studio for small models such as Muse Glimmer 30B. Large models like Kimi K3 or Mistral Large 4 need data-centre GPUs to self-host.