Thinking Machines发布Inkling 2 Small，AA综合评分达31分
Thinking Machines发布Inkling 2 Small，Artificial Analysis综合评分达31分，Agent任务和编程测试成绩较上一代提升。
Loading…
bibibi.ccbibibi.ccLoading…
Connect scattered updates into a story you can understand.
Thinking Machines发布Inkling 2 Small，Artificial Analysis综合评分达31分，Agent任务和编程测试成绩较上一代提升。
Dongcha Beating AI reports that Meituan has released the LongCat-2.5 official version, focusing on improved visual programming and complex Agent task capabilities, and has opened online trial and API。
DeepSeek Harness’s official team published an article introducing an Router architecture consisting of a dual-layer structure of Worker and Watchdog event-driven operation and Super-agent among other components. The design aims to improve the continuity, proactivity, and task persistence of Agent systems.
DiffuSpace released and open-sourced a diffusion model based on a modification of Qwen3.6-35B-A3B Clover-35B-A3B-Preview. Its disclosed internal tests reportedly showed that the model reached GB300 on a single GPU 680 Token/s。
BOSS Nanbeige Laboratory, under Direct Hire, has released and open-sourced Nanbeige4.3-33B, for programming and office work Agent, supporting 512K context and using the Apache 2.0 license. The team plans to distill its capabilities into subsequent 3B models.
BlockBeats stated that MiniMax open-sourced M3.1 Flash model weights, claiming that its Agent task costs are reduced by 70%and generation speed is close to 3 times. The model uses a 428000000000-parameter MoE architecture, activating approximately 23000000000 parameters per generation.
Meta Muse Spark Open, released by 39, exceeded Nvidia Nemotron by 3.5 Super VL points on the Artificial Analysis Intelligence Index, but Agent lagged behind several Chinese open-source models in combined performance on the
Anthropic opens the OpenTelemetry tracing public beta for Claude Platform, allowing users to view Agent call duration, errors, and Token usage, and send execution records to external monitoring platforms.
AI Infrastructure company Modal says it completed 750000000 USD Series D financing, led by Accel, with a post-money valuation of 15750000000 USD. The company also disclosed that annualized revenue had exceeded 1000000000 USD, while annualized revenue from the Sandboxes business reached 300000000 USD。
Liquid AI released a new local mixture-of-experts model, emphasizing Agent reasoning and tool calling. The model leads Qwen in some Agent tests but performs more weakly in programming tests.
After establishing its security and alignment department, ByteDance Seed launched global recruitment, planning to hire 10 researchers, focusing on Agent privilege escalation, model attack and defense, and the safety of high-risk capabilities.
A review of 51 public usage records REA shows that most independent attempts completed only partial functionality, while complex software remains difficult to reproduce in full.