Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

The best AI coding models in 2026

The best AI coding models right now are GLM-5.3, Hy4 preview and Qwen3.6, out of 16 we have reviewed. Each is ranked by our rating, with its price, our verdict and who it suits.

By ReviewRamble editors · Updated Nov 14, 2026 · How we review

How we ranked: by our rating out of 5, from reviews built on each product's own pricing, documentation and policies plus independent benchmarks where they exist, each checked on the date shown in the review. We take no affiliate commission. Models whose maker has retired the hosted API stay listed for their open weights and are marked.

  1. 1

    Open-weight coding within a few points of the closed frontier, under a licence almost nobody needs to read twice.

    Best for: Developers and companies who want a capable open coding and agent model with minimal licence restrictions.

    Price: Open weights under a modified MIT licence. Z.ai API: $1.40 input / $0.26 cached / $4.40 output per 1M tokens; GLM-5.3-Flash $0.15 / $0.50.

    Read our GLM-5.3 review
    4.4/5
  2. 2

    A genuinely open (Apache 2.0) coding model that trades blows with Kimi K3 at a third of its size, held back by preview-stage habits and no independent score yet.

    Best for: Teams building coding agents who want a permissively licensed open model they can self-host on one server.

    Price: Open weights under Apache 2.0. Tencent Cloud TokenHub: ¥6 input / ¥18 output per 1M tokens, or $0.834 / $2.501 internationally, cached input $0.042. OpenRouter from about $0.75 / $2.25. Hy3: $0.13 / $0.53.

    Read our Hy4 preview review
    4.2/5
  3. 3

    Two strong Apache-licensed coders where the surprise is that the smaller dense 27B beats its larger mixture-of-experts sibling on almost every test.

    Best for: Teams who want a proven, permissively licensed coding model and are choosing between a dense build for quality and an MoE build for throughput.

    Price: Free weights under Apache 2.0 for both sizes. Hosted free on Qwen Chat; you pay the GPU to self-host or a third-party host's per-token rate.

    Read our Qwen3.6 review
    4.2/5
  4. 4

    A capable open coding model that clearly beats its own predecessor and still trails the closed frontier, with an awkwardly specialised "always thinking" design.

    Best for: Teams self-hosting the Kimi K2 line who want the strongest open, agentic coding model in that family and can run a 1T-parameter INT4 model.

    Price: Free weights under a Modified MIT License. Hosted on Moonshot's Kimi API and Kimi Code plan; per-token pricing is not stated on the model card.

    Read our Kimi K2.7 Code review
    4.1/5
  5. 5

    Near-frontier agentic coding from a model small enough for one workstation, at a price that rounds to zero.

    Best for: Teams that want a cheap, permissively licensed open coding model they can run on a single workstation or call for cents.

    Price: Open weights under OpenMDW-1.1. Launch OpenRouter price: $0.10 input / $0.20 output / $0.01 cache read per 1M tokens; free 256K-context endpoint.

    Read our Laguna S 2.1 review
    4.1/5
  6. 6

    Still a sensible, cheap choice for inline autocomplete, but it is fifteen months old and the wrong tool for anything beyond completing the next few lines.

    Best for: Developers and IDE vendors who want cheap, fast inline code completion through an API.

    Price: Mistral API: $0.30 input / $0.90 output per 1M tokens; batch $0.15 / $0.45. Enterprise self-hosting by arrangement.

    Read our Codestral 25.08 review
    3.9/5
  7. 7

    A 72.1 GB build that keeps 94.8% of DeepSeek-V4.1-Flash's SWE-bench score on the lab's own test, paid for with weaker prose, weaker non-English output and another 110.6 GB of disk.

    Best for: Developers with 64 GB or more of RAM, a fast NVMe drive and one or two large GPUs who want DeepSeek-V4.1-Flash for coding agents without sending code to an API.

    Price: Free MIT-licensed weights, 182.72 GB of GGUF files including the Engram table; the cost is the hardware. The full base model on DeepSeek's API costs $0.15 in / $0.60 out per 1M tokens off-peak.

    Read our DeepSeek-V4.1-Flash-1017 GSQ-RCO Coder (GGUF) review
    3.9/5
  8. 8

    A coding model genuinely sized for a 48 GB laptop, from a well-funded lab, with every headline score still Recursive's own.

    Best for: Developers who want a fast coding agent model that runs locally on a Mac or a single GPU.

    Price: Free open weights under Apache 2.0; free, rate-limited endpoint on OpenRouter. Paid Flint API for select partners, price not published.

    Read our Flint review
    3.9/5
  9. 9

    Ripple-65B makes a credible case that a 65B dense model can match American open models eight to fifteen times its size at agentic coding, but every score so far is Ricursive's own.

    Best for: Platform teams with two B200s, or a DGX Spark, who want an Apache 2.0 coding-agent model they can host and fine-tune themselves.

    Price: Free open weights under Apache 2.0. Hosted on OpenRouter (GMICloud) at $0.20 per 1M input tokens and $0.60 per 1M output tokens.

    Read our Ripple-65B Preview review
    3.9/5
  10. 10

    A fast little coding agent for your own GPU that fixes routine bugs, though Qwen3.5-9B still edges it on the hard ones.

    Best for: Teams building coding agents that need fast, cheap sub-agents on their own hardware.

    Price: Free weights under Apache 2.0. No hosted API; GGUF builds from 7.03 GB run locally.

    Read our Mellum2.1 Thinking review
    3.8/5
  11. 11

    On its makers' numbers Nanbeige4.3-33B is the strongest open coding agent near 30B, but no one outside the lab has confirmed that yet, and being dense makes it heavy to serve.

    Best for: Teams that self-host and want an Apache-licensed open coding agent, or a base model to post-train, and can verify its claims on their own tasks.

    Price: Open weights under Apache 2.0; no official hosted API.

    Read our Nanbeige4.3-33B review
    3.8/5
  12. 12

    A 4B model that resolves six in ten SWE-bench Verified issues is a real research result, but every number is Microsoft's own and the harness is built for a lab, not a laptop.

    Best for: Researchers and platform teams studying small coding agents, or running sandboxed bug-fix agents on their own GPUs where every patch is reviewed by a person.

    Price: Free weights (tagged MIT on Hugging Face; the card's table says Apache 2.0). No hosted API; you serve it yourself, for example with SGLang.

    Read our FrogNano-4B-2609 review
    3.7/5
  13. 13

    MAI-Code-1.1-Flash is a sensible cheap default inside Copilot, but it trails DeepSeek's Flash models on both price and Terminal-Bench and is barely available anywhere else.

    Best for: GitHub Copilot users and teams who want a cheap, fast default for routine edits and CLI tasks without leaving Copilot.

    Price: Provider list price $0.20 per 1M input tokens, $0.02 cached and $1.20 output, 73% below MAI-Code-1-Flash; billed through GitHub Copilot usage, or at a 0.25x premium request multiplier on annual Copilot plans. Local calls carry no inference charge.

    Read our MAI-Code-1.1-Flash review
    3.6/5
  14. 14

    Striking coding scores from a lab with no public names, and a one-GPU pitch that holds only at 4-bit.

    Best for: Teams serving open-weight coding agents on a single large GPU who can run their own evaluations.

    Price: Free open weights under MIT. You pay only for your own hardware or hosting.

    Read our Ornith-1.5-122B review
    3.6/5
  15. 15

    Hosted API retired Mar 31, 2026; open weights still available

    Devstral Small 2 still runs well on one gaming GPU, but Mistral has abandoned the line and nobody now sells it as a service.

    Best for: Developers who want a self-hosted coding agent on one 24GB GPU or a 32GB Mac and are content to maintain a model its maker no longer develops.

    Price: Free weights under Apache 2.0. Mistral's API version was retired on March 31, 2026, and Artificial Analysis lists no public API.

    Read our Devstral Small 2 review
    3.4/5
  16. 16

    Hosted API retired Jul 31, 2026; open weights still available

    A solid coding model for late 2025 that its own maker has retired; new projects belong on Mistral Medium 3.5 or a newer open model.

    Best for: Existing self-hosted users who are not ready to migrate yet.

    Price: Open weights under a modified MIT licence (not for companies over $20M monthly revenue). Retired from Mistral's API on July 31, 2026.

    Read our Devstral 2 review
    3.1/5

Sources

The main official source behind each ranked review; every review lists its full set.

More guides