A comparison of DeepSeek-V4 Flash 0731 and GPT-5.6 Luna on the DeepSWE benchmark reveals that while GPT-5.6 Luna is the stronger engineer, a cascade strategy using both models achieves higher accuracy at lower cost.
Moonshot AI has released Kimi K3, a 2.8-trillion-parameter open-weights model designed for long-horizon coding and deep reasoning. It is the world's first open-source model in the 3-trillion-parameter class and supports a 1 million token context window.
Together AI has announced a strategic partnership with Moonshot AI to serve as the launch platform for Moonshot's open-weight models, starting with Kimi K3. This collaboration provides developers with day-zero access to frontier releases through Together AI’s US-hosted infrastructure and tooling.
A comparison of Kimi K3 and GPT-5.6 Sol on the DeepSWE benchmark reveals that while GPT-5.6 Sol leads in single-shot quality (72.7% vs 68.5%), Kimi K3 achieves higher pass@k scores for k > 1 at a significantly lower cost.
Kimi K3, an open-weight model from Moonshot AI, achieves performance comparable to Anthropic's Claude Fable 5 on the DeepSWE benchmark while costing significantly less per task. In a July 16, 2026 evaluation, Kimi K3 reached 68.5% pass@1 compared to Fable 5's 69.9%, but surpassed it in multi-attempt scenarios and offered superior cost efficiency.
Thinking Machines Lab has released Inkling, a new multimodal mixture-of-experts model designed for token-efficient reasoning and native understanding of text, image, and audio inputs. Together AI is making the model available on its inference platform with day-zero access.
Together AI has added autoscaling capabilities to its Dedicated Model Inference platform, allowing deployments to scale based on metrics the inference engine understands, such as in-flight requests, time-to-first-token (TTFT), and GPU utilization.
ThunderAgent is a system that introduces a program abstraction for agentic LLM request scheduling to address KV cache thrashing in high-concurrency synthetic data generation. By treating agent workflows as schedulable programs rather than independent requests, it significantly improves throughput and reduces latency.
Together AI details the architecture of its Dedicated Model Inference platform, which links endpoints, deployments, and immutable configs through a capacity-aware traffic split. This system routes requests based on effective capacity (weight multiplied by ready replicas) rather than fixed percentages, enabling features like zero-downtime rollouts and A/B testing.
Together has released a significant update to its inference platform, introducing "Dedicated Model Inference" designed to give users complete control over performance, cost, and quality for open-weight models. The update also announces a closed beta for custom training capabilities, including full-weight and LoRA reinforcement learning.
Together AI and Y Combinator have announced a partnership to provide the first dedicated GPU cluster exclusively for Y Combinator's portfolio of AI-native startups. This initiative aims to solve the critical bottleneck of compute access by offering flexible, cost-effective infrastructure for training and inference without requiring long-term commitments.
Together AI details the specific architectural requirements needed to achieve different reliability tiers for GPU inference, arguing that standard SLA numbers often obscure the actual failure domains covered.
Together has updated its GPU Cluster platform with new reliability and operational control features designed for large-scale training and inference workloads. The changes include a rebuilt Slurm-on-Kubernetes stack, automated node repair mechanisms, and enhanced access management tools.