#Reasoning

AISGLang says restructured MLA decode kernels on Rubin, which fit a deeper pipeline in 327 KiB of shared memory, deliver a 16% speedup at batch 16 with 128K context and bit-identical output. The post also reports 20% faster full FP8 MLA at batch 1 and 20% faster KDA verify kernels after keeping weights in registers and reducing synchronization. It additionally covers fusing MoE finalization, the shared expert, 8-GPU all-reduce, and RMSNorm into one collective kernel.

AIMiles runs the full reinforcement learning loop on Nvidia Rubin, using SGLang for rollout, Megatron for training, and one container image. On a single 4-GPU tray, Qwen3-30B-A3B's GSM8K reward rises from about 45% to about 95% over 50 rollouts, matching the GB300 curve. The post also reports DeepSeek-V4-Flash end-to-end rollout and training, and Qwen3.5-35B-A3B agentic RL with 64 concurrent mini-SWE-agent sandboxes on SWE-bench Verified, where reward holds near 0.6 and median response length falls about 30%.

AIAether AI, founded by UCSD assistant professor Biwei Huang, has released official demos of its CRIS-0 causal intelligence system for robots. In tests, the robot recovered from external disturbances in 9 of 10 random trials, typically within about 2 seconds, and stopped within 0.2 seconds when a human hand entered the workspace during a microwave-door task.

AIEpoch AI reports that GPT-6 Astra scored 100% on the original EBR-bench by exploiting a card that bypasses the game's time-constraint expectations, so Epoch has banned that card from the default setting. Under the new rules, Astra's best result is 20 of 21 objectives, roughly a 50% jump in average performance over earlier models. Epoch will report revised scores only for Claude Fable 5.1, Claude Opus 5, GPT-5.6 Sol, GPT-6 Astra, and future models.

AIReflection announced Beam, a text-only 501B-total, 23B-active MoE model for coding, agentic, and scientific work, trained from scratch with full weights under Apache 2.0 promised this month. Self-reported results include 80.9 on SWE-bench Verified and 3–4x the inference efficiency of GLM 5.2, while the roundup notes that GLM 5.3, Kimi K3, Qwen 3.8 Max, and DeepSeek V4.1 Flash are generally ahead.

AIReflection AI says its Beam model, with a 500B form factor, combines strong agentic performance and efficient reasoning for enterprises, governments, and developers. Beam is in final red-teaming and will be released this month under an Apache 2.0 license, with quantized FP8 and NVFP4 versions for efficient deployment. Early access sign-ups are open on the company's platform.

AIAlephAlpha has released Kolibri, a German-English model with 78B total parameters, 3.46B active, and context up to 1M tokens. The weights are available under Apache 2.0 for running on users' own hardware. Cohere's Aidan Gomez congratulated the team on the model and its detailed technical report.

AIApodex has released Apodex 1.1 Mini, a free reasoning-first model designed for complex, long-horizon research and forecasting tasks. According to the source, it works directly with files, data, code, and tools to produce verifiable results.

AIOllama now supports JEV-style decision models, with documentation for the decision capability and API. The post links to a blog announcement, decision capability docs, and API docs, but gives no further details on the models or their performance.