Alternatives to Kolibri-1

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.

Fireworks Research rebuilt Kimi K3 to think less: same benchmark quality with roughly 40 percent fewer tokens, a 1M context, and $3 input per million tokens.

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.

Tencent's next-generation open-weight MoE flagship: 770B total params, 49B activated per token, 1M-token context. Apache 2.0, Gated DSA attention, and blind-eval scores edging out GLM 5.3 and Kimi K3.

Meituan's open-weight MoE model: 1.6T total / ~48B active params, 1M context, MIT. Strong on coding and agentic tasks (2026-08-25).

NVIDIA's efficient open-weight 30B MoE hybrid model with 3B active parameters, 1M-token context, and single-GPU deployment for local reasoning and coding.

Bespoke Labs' open-weight 9B model and training recipe that turn Jev-style decisions into single-token choices with probabilities, runnable locally.

Google DeepMind's open-weight 25.2B MoE model with 3.8B active parameters, 256K context, multimodal input, and a commercially permissive Apache 2.0 license.

K2-Horizon-7B is an open-weight IFM language model with a 512K context window, tool calling, and released training checkpoints.

Xiaomi's MiMo-V2.6 open-weights series pairs a 1.02T-parameter omni MoE with a 309B Flash tier, both MIT licensed with 1M-token context.

BAAI's 27B open-weight agent model for long-horizon self-improvement: reflect, measure, and revise across coding, ML engineering, and deep research with a 262K context.

Agnes-3.0-Flash Preview is an Apache-2.0 33B multimodal open-weight checkpoint with 262k context, not the proprietary API model.

Swift-Qwen3.8-27B is UkisAI's reasoning-efficient Qwen3.8-27B finetune that cuts thinking tokens about 58% with under 1% score loss.

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.