AI Coding Live News

The SGLang team released v0.5.22 featuring 950 PRs from 245 contributors: DeepSeek-V4.1 delivers up to 2.3x more throughput per GPU on NVIDIA H200 and 1.9x on AMD MI350X/MI355X via DSpark. The new InstantTensor loader cuts model loading time by 10x. B200 decodes Qwen3.8-Flash-Next 12% faster, while B300 speeds up GLM-5.3-Flash by 10% and yields 1.85x decode throughput for MiMo-V2.6-Pro. SGLang also adds native Apple Silicon PyTorch support and ComfyUI integration.

Zoom

On 2026-10-14 Inworld AI opened Inworld Inference to every developer, positioning it as an inference service for consumer apps (companions, tutors, voice, interactive fiction). Instead of listing thousands of models it builds a per-model serving stack for open models such as Gemma 4, GLM 5.3 Flash, DeepSeek and MiMo: draft-and-verify (speculative) decoding, a rewritten split-KV verify attention kernel, KV-aware routing and eval-gated low precision. Vendor-reported: Gemma 4 26B p95 first token of 0.31 s, 7.8x faster than the median OpenRouter provider; up to 90% of prompts from cache; 99.999% success over 20M+ GLM 5.3 Flash requests in 30 days. The API is OpenAI-compatible with Flex/Default/Priority/Fast tiers per request.

Zoom

Vals AI audited all 2,698 coding tasks in the RL environments Xiaomi open-sourced with MiMo v2.6: in 1,795 (67%) the reference fix survives as unreachable Git objects. MiMo v2.6 finds and copies it, writes its own pack-file parser when git commands are blocked, and uses file mtimes when Git is removed.

Zoom

Reinforcement learning (RL) pipelines for code agents typically rely on binary test suite outcomes for reward assignment. Under standard Group Relative Policy Optimization (GRPO), all test-passing rollouts within a sampled group receive identical advantage values, completely ignoring code quality, extraneous diff modifications, and algorithmic elegance. This induces severe trajectory bloat and training instability. Researchers introduce GAGAR (Groupwise Agentic Grading and Advantage Redistribution, arXiv:2609.32577). In a shared workspace, an agentic evaluator jointly inspects all passing candidates in a group, ranking them by architectural clarity and minimal intervention. A sum-preserving redistribution reallocates advantage weights toward superior implementations. Validated at industrial scale on MiMo-V2.6-Flash (310B) and MiMo-V2.6-Pro (1.02T parameters), GAGAR curtails trajectory length explosion while boosting code agent benchmark performance and optimization stability.

Zoom

SGLang, a leading open-source high-throughput inference engine for large models, officially releases milestone version v0.5.21, merging 779 PRs across 227 contributors. The release introduces dynamic on-the-fly role switching between Prefill and Decode (PD) instances without restarting, an upgraded high-performance Rust core for the default prefix cache, and 22% faster first-token generation for DeepSeek-V4.1 on long prompts. It also debuts the /v1/decisions and /v1/score APIs, alongside complete support for next-gen models including MiMo-V2.6, DiffusionGemma, and GLM-5.3-Flash on AMD MI355X.

Zoom

Reinforcement learning for code agents typically relies on executable tests for binary rewards, leading algorithms like GRPO to assign identical scalar advantages to all passing rollouts regardless of code bloat, extraneous modifications, or architectural hygiene. Researchers from ByteDance, Peking University, and Tsinghua introduced GAGAR, a quality-aware advantage redistribution framework. By staging rollouts in a shared workspace where an SFT agentic grader ranks passing candidates and dynamically shifts credit toward clean solutions, GAGAR successfully stabilizes code agent RL on 310B Flash and 1.02T Pro models.

Zoom

llama.cpp v0.5.0 (2026-09-23) accelerates CUDA conv2d via implicit GEMM, adds Metal MoE and SSM_CONV fusion, lets the server bind multi-address/--host lists and UNIX sockets, and adds HRM-Text (DFM Mimir 1B), MiMo-V2.6, and HunyuanOCR (DFlash) conversion plus ggml 0.25.0 backends.

Zoom

@Zaddyzaddy published @BugBunny_ai VulnPR-100 results: on 100 PRs with known vulns (up to one hour each), MiMo-V2.6-Pro found 46 for $8.27 — best open-weight score, matching GPT-6 Astra at roughly 22× lower review cost. MiMo-V2.6-Flash found 41 for $3.47. That is about $0.18 per vuln found for Pro and $0.08 for Flash.

Zoom

@arena reports Xiaomi MiMo-V2.6-Pro debuted on Code Arena: WebDev around #10 overall and about #3 among MIT-licensed open-weights models, scoring 1628 AutoEval points—tying Claude Fable 5 (High) and just ahead of Hy4-preview (1624). That is a +153 jump from MiMo-V2.5-Pro (1475). Live human votes are still coming in.

Zoom

@ArtificialAnlys reports Xiaomi released open-weights MiMo-V2.6-Pro: Artificial Analysis Intelligence Index 46 (up from 26 on V2.5-Pro), now the top open-weights model, landing on the intelligence-vs-cost Pareto frontier at about $0.13 per Index task with unchanged token prices ($0.435/M input with 99% cache-hit discount, $0.87/M output). It is a 1.02T-total / 42B-active MoE.

Zoom