NeoHorse-1-4B is a 4B causal language model released by TokenRhythm under the apache-2.0 license. It is a fine-tune of Qwen/Qwen3.5-4B, post-trained for text-based agent harnesses, tool use, coding, and instruction following, and described as an initial prototype toward recursive self-improvement. The release contains language-model weights only, is repackaged for text-only inference, and omits vision weights; repackaging changed configuration and tensor key names without changing fine-tuned tensor values.
Model size
4.21B
Format
FP16 weights
Context
262,144
Freshness 86 out of 100. Repository age: 40 days old, Sep 5, 2026. Measures age, not quality.Freshness86/10040 days old
Popularity 62 out of 100. 15,703 downloads in 30 days; 2,550 lifetime likes. Checked Oct 15, 2026. Measures adoption, not quality.Popularity62/10015.7K downloads / 30d
NeoHorse-1-9B is a 9B causal language model released by TokenRhythm under the Apache-2.0 license. It is a fine-tuned derivative of Qwen/Qwen3.5-9B, post-trained for text-based agent harnesses, tool use, coding, reasoning, and instruction following. The release contains language-model weights only and is repackaged for text-only inference; vision weights are not included. The official model card describes it as an initial prototype toward recursive self-improvement and mentions a routing harness that assigns tasks to a heterogeneous model pool, records tool interactions and outcomes, estimates capability demand, and uses capability-level feedback to shape the next training mixture.
Model size
8.95B
Format
FP16 weights
Context
262,144
Freshness 86 out of 100. Repository age: 40 days old, Sep 5, 2026. Measures age, not quality.Freshness86/10040 days old
Popularity 56 out of 100. 6,616 downloads in 30 days; 1,029 lifetime likes. Checked Oct 15, 2026. Measures adoption, not quality.Popularity56/1006.6K downloads / 30d