Kimi K2.5
kimi-k2-5-ac79b1bf·7 events·first seen 27d agoAliases: Kimi K2.5, Kimi-K2.5, Kimi-k2.5
Merged from
Kimi-K2.5
Co-occurring entities
More like this (12)
Recent events (7)
Data Points: Cursor Composer 2.5, Gemini 3.5 Flash, Antigravity 2.0, Omni Flash, AI Search, and Corti Symphony
This edition covers several notable AI product and model releases: Cursor shipped Composer 2.5 (built on Kimi K2.5) scoring 79.8% on SWE-Bench Multilingual at significantly lower cost than frontier competitors; Google released Gemini 3.5 Flash with claimed 4x speed advantage and launched Antigravity 2.0 as an agent-first desktop app replacing its IDE; Google also introduced Gemini Omni Flash for multimodal video generation and overhauled its search interface with Gemini 3.5. Additionally, Copenhagen-based Corti launched Symphony for Speech-to-Text achieving 1.4% word error rate on medical terminology versus 17-19% for generalist models.
Cursor's Composer 2.5 rivals GPT-5.5 and Claude Opus 4.7 on coding benchmarks at lower cost
Cursor released Composer 2.5, a specialized agentic coding model built on Moonshot's Kimi K2.5 open weights with additional pretraining and reinforcement learning fine-tuning tailored to Cursor's own CLI harness. The model ranks third on the Artificial Analysis Coding Agent Index behind Claude Opus 4.7 and GPT-5.5 at max reasoning, but significantly undercuts them on cost ($0.44 vs $4.14 per task) and speed (6.7 vs 17.7 minutes). The training approach—co-optimizing model and harness together using synthetic tasks, text feedback during RL, and 25x more synthetic data than Composer 2—illustrates a specialist model strategy that challenges the dominance of generalist frontier models in coding workflows.
ETCHR: Decoupled Image Editing for Visual Chain-of-Thought Reasoning in MLLMs
ETCHR introduces a question-conditioned, reasoning-aware image editing model that decouples visual transformation from downstream understanding in multimodal LLMs. It addresses two identified gaps—language-side (mapping abstract questions to visual edits) and generation-side (edit quality degrading with reasoning depth)—via a two-stage training recipe combining supervised fine-tuning on edit trajectories and VLM-derived reward signals. Because the editor is decoupled, it plugs into arbitrary MLLMs without retraining, yielding Pass@1 gains of roughly +4.6 to +5.5 points across five task families when paired with Qwen3-VL-8B, Gemini-3.1-Flash-Lite, and Kimi K2.5. The work advances the 'think with images' paradigm beyond fixed toolkits and unified multimodal approaches.
Mistral Releases Leanstral: First Open-Source Code Agent for Lean 4 Formal Verification
Mistral AI has released Leanstral, an open-source code agent built on a sparse 120B/6B-active-parameter architecture, designed specifically for formal proof engineering in Lean 4. The model targets realistic proof engineering workflows rather than isolated math competition problems, and is benchmarked on FLTEval, a new evaluation suite tied to the Fermat's Last Theorem formalization project. Leanstral is released under Apache 2.0 with a free API endpoint and MCP support, and demonstrates competitive performance against Claude Sonnet 4.6 at roughly 1/15th the cost. The release positions formal verification as a scalable alternative to human code review for high-stakes software and mathematics.
Nvidia releases Nemotron 3 Super 120B-A12B open-weights model with hybrid Mamba-2/MoE architecture
Nvidia released Nemotron 3 Super 120B-A12B, an open-weights LLM with a hybrid Mamba-2/transformer/MoE architecture that activates only 12B parameters per token and supports up to 1 million token context. The model claims the fastest inference speed in its size class at 442 tokens/second and leads open-weights models on PinchBench agentic task evaluation, outperforming larger models including Kimi K2.5 (1T parameters). Nvidia is releasing weights, training data, and recipes under a permissive commercial license, and plans a $26B five-year investment in open-weights models — framed partly as a strategic response to Chinese labs building capable open-weights models on non-Nvidia hardware.
Meta Introduces Muse Spark: First Closed-Weights Model from Superintelligence Labs
Meta released Muse Spark, its first AI model in roughly a year and the debut product of its Superintelligence Labs, marking a significant departure from its open-weights Llama strategy. The natively multimodal reasoning model supports tool use and multi-agent orchestration, achieves fourth place on the Artificial Analysis Intelligence Index, and claims notable token efficiency—matching Llama 4 Maverick with over 10x less training compute. Meta withheld parameter count, architecture, and training details, positioning Muse Spark as a closed commercial product competing with OpenAI, Google, and Anthropic. The release introduces 'thought compression' via RL and a parallel multi-agent 'contemplating' mode, while showing gaps in coding and agentic benchmarks.
HyperTool: Unified executable MCP-style interface reduces step-wise tool call overhead for LLM agents
HyperTool introduces a unified executable interface that allows LLM agents to invoke multiple tool calls within a single code block, hiding intermediate dataflow from the main reasoning trace. This addresses an 'execution-granularity mismatch' where step-wise atomic tool calls waste context and force models to manage low-level operations. On the MCP-Universe benchmark, HyperTool more than doubles accuracy for Qwen3-32B (15.69% → 35.29%) and Qwen3-8B (9.93% → 33.33%), outperforming GPT-OSS and Kimi-k2.5.