Alternatives to Recursive Flint

Reflection AI's Apache-2.0 open-weight 501B MoE for coding and long-horizon agent work, with 23B active parameters and 1M context.

Meta's open-weight 30B agentic multimodal model distilled from Muse Spark, designed to run always-on coding and tool-use agents locally on a single consumer.

Zhipu's June 2026 744B MoE coding model with 1M context and MIT weights. Successor GLM-5.3 is now the flagship.

Open-weight dense reasoning LLM from IBM: 30B parameters, Apache 2.0, native 128K context extendable to 512K, built-in chain-of-thought and reasoning-augmented tool calling.

Kwaipilot's open-weight agentic coding model: 35B MoE, 3B active per token, Qwen3.6 base, Apache 2.0, 262K context, top PinchBench tool-use score.

OpenBMB MiniCPM5-2B is a dense 2B on-device LLM with 131k context, Apache-2.0 weights, and cookbooks for vLLM, SGLang, and llama.cpp.

Von is an Apache-2.0 395M System One decision model that answers typed questions locally in about 18 ms on Apple MPS, as an open answer to TypeSafe Jev.

DeepSeek flagship 1.6T MoE with 49B active, 1M context, MIT weights. API $0.29/$1.20 cached $0.03.

Zhipu coding model on GLM Coding Plan. Token API not on the public table yet. Not $18-only.

Poolside's open-weight 118B MoE coding model with 8B active parameters, a 1M-token context window, native interleaved reasoning, and an OpenMDW-1.1 license.

Liquid AI's open-weight 24B reasoning MoE with 2.45B active parameters, 128K context, and tool calling for local agents.

Meta's open-weight, multimodal MoE for coding and agents, with 816B total parameters, 42B active parameters, and 1M-token context.

Agentic coding MoE. Not a lab release.

inclusionAI MIT hybrid-linear MoE: 124B total, 5.1B active.