Deepcoder 14b New Opensource Coding Model Beats 03 Mini Tested In this video, we explore deepcoder 14b, the latest open source coding model released by together ai. In this work, we introduce deepcoder 14b preview, a 14b model reaching o3 mini performance with 60.6% pass@1 accuracy for livecodebench. to achieve this, we curated high quality, verifiable coding data and introduced algorithmic and system optimizations for effective rl training.
Deepcoder 14b New Opensource Coding Model Beats Deepseek R1 O1 O3 Deepcoder 14b preview is a code reasoning llm fine tuned from deepseek r1 distilled qwen 14b using distributed reinforcement learning (rl) to scale up to long context lengths. Explore the training and benchmarks of open source coding model deepcoder 14b and see how it compares with the performance of o3 mini & o1. A 14 billion parameter code generation and reasoning model jointly trained by agentica and together ai using reinforcement learning with verifiable rewards (rl vr). We introduce deepcoder 14b preview, trained on 24k verifiable coding problems over 2.5 weeks on 32 h100s, reaching—and even surpassing—openai’s o3 mini on various coding benchmarks.
Deepcoder 14b Reaches Openai S O3 Mini Level Accuracy In Livecodebench A 14 billion parameter code generation and reasoning model jointly trained by agentica and together ai using reinforcement learning with verifiable rewards (rl vr). We introduce deepcoder 14b preview, trained on 24k verifiable coding problems over 2.5 weeks on 32 h100s, reaching—and even surpassing—openai’s o3 mini on various coding benchmarks. The report highlights open source contributions, reinforcement learning training optimizations, and significant model performance improvements across coding, multimodal, and image generation domains. Through a joint collaboration between the agentica team and together ai, we release deepcoder 14b preview, a code reasoning model finetuned from deepseek r1 distilled qwen 14b via distributed rl. Together, ai and agentica have released the deepcoder 14b preview, a new open source model capable of coding performance at the same level as the very best commercial systems. Deepcoder 14b represents a potential turning point in this landscape, offering performance that rivals openai ’s o3 mini while maintaining complete transparency from model weights and training data to reinforcement learning code and detailed logs.
New Model Drop Deepcoder 14b Preview By Agentica Moshe G Posted On The report highlights open source contributions, reinforcement learning training optimizations, and significant model performance improvements across coding, multimodal, and image generation domains. Through a joint collaboration between the agentica team and together ai, we release deepcoder 14b preview, a code reasoning model finetuned from deepseek r1 distilled qwen 14b via distributed rl. Together, ai and agentica have released the deepcoder 14b preview, a new open source model capable of coding performance at the same level as the very best commercial systems. Deepcoder 14b represents a potential turning point in this landscape, offering performance that rivals openai ’s o3 mini while maintaining complete transparency from model weights and training data to reinforcement learning code and detailed logs.
Deepcoder A Fully Open Source 14b Coder At O3 Mini Level Together, ai and agentica have released the deepcoder 14b preview, a new open source model capable of coding performance at the same level as the very best commercial systems. Deepcoder 14b represents a potential turning point in this landscape, offering performance that rivals openai ’s o3 mini while maintaining complete transparency from model weights and training data to reinforcement learning code and detailed logs.
Comments are closed.