Tag
Jeeves is a 9B parameter reasoning-enhanced Jev-like decision model that uses a diffusion drafter, achieving higher accuracy on out-of-domain benchmarks compared to previous models like Kev and Jev.
A $500 RL fine-tune of a 9B open-source model achieves 87% catalog review quality at $0.50 per 1000 listings, significantly outperforming frontier models like GPT-4 and Claude which cost $19-$172 and achieve only 70-76% quality.
Describes a personal project where an AI agent powered by a 9B model running via Ollama serves as a DJ to select music.
Qwythos-9B is a 9B super reasoning model based on the Qwen3.5-9B base, further trained with 500 million+ Claude Mythos reasoning traces. It natively supports 1M long context and tool calling, designed for local deployment by security researchers. It significantly outperforms the original base model on MMLU and mathematical reasoning.
Discussion about Alibaba's potential open-weights release of Qwen 3.7 9B and seeking alternatives to Qwen 3.5 9B for local deployment.
A commentator discusses the performance of a 9B model on coding benchmarks, noting that while it beats its base on SWE-bench (69 vs 53), the advantage narrows on behavioral and long-horizon tests, suggesting limited gains outside benchmark distributions.
Ornith-9B demonstrates that RL training at 9B parameters primarily buys efficiency, achieving same answers with ~56% of the tokens and twice the speed of its base model, offering real cost savings for per-token payment.
The author evaluates Ornith-9B against its base Qwen3.5-9B, finding that RL post-training improves token efficiency and sustained coding coherence but sacrifices single-turn judgment and robustness to misleading inputs, making it a narrower upgrade at 9B compared to the 35B version.
Trained a 9B model (Qwythos 9B) on Claude Mythos traces, achieving strong results in bug-finding, SQL, and function calling while running locally on an A6000 with llama.cpp.
Ornith-1.0-9B is a new 9B parameter AI model optimized for 8-12GB GPUs, achieving strong performance on agentic coding benchmarks, matching or surpassing models 2-3x its size.
Ai2 and the University of Washington released a paper titled Tmax, proposing the strongest open-source terminal agent RL training recipe to date. A 9B parameter model outperforms larger models on Terminal-Bench 2.0, with the key being low-cost generation of vast amounts of verifiable training data, not model size or algorithm.
A new 9B fine-tuned model called Qwopus3.5-9B-Coder is released, optimized for tool calling and agentic coding workflows, achieving strong SWE-bench and HermesAgent-20 scores while running on affordable hardware.