AI Coding Live News — Real-Time 𝕏/Twitter Updates & Breakthroughs

👀161

🔥90

Moonshot launches Kimi K3.1 Pro: 2.8T-total / 104B-active open-weight MoE with 1M context, 65.7% on Terminal-Bench 4.0, API at $2/$10, weights by Nov 4

On 2026-10-23 (17:15 UTC) the official @Kimi_Moonshot account launched Kimi K3.1 Pro, alongside K3.1 Pro Highspeed (2-3x faster). Both are live on Kimi.ai, Kimi Work, Kimi Code and the Kimi API, and full weights are promised by 2026-11-04. The official tech blog describes it as a post-trained version of K3 with much larger multi-domain joint RL: a 2.8T-total / 104B-active MoE, 1M-token context, multimodal. Vendor-reported scores: 65.7% on Terminal-Bench 4.0 (GPT-6 Astra 58.2%, Claude Opus 5.5 66.4%) and 92.7% on BenchCAD with tools. The Kimi API pricing page lists kimi-k3.1-pro at $2 input, $10 output and $0.20 cached input per million tokens, down from K3 at $3/$15.

Key Takeaways

2.8T-total / 104B-active MoE, 1M-token context, multimodal; full weights promised by 2026-11-04 (official blog)
Vendor-reported: 65.7% on Terminal-Bench 4.0, ahead of GPT-6 Astra (58.2%) and just behind Claude Opus 5.5 (66.4%); 61.2% on the internal Kimi Code Bench 3.0 Long-horizon (K3: 55.95%)
API pricing: kimi-k3.1-pro at $2 input / $10 output / $0.20 cached input per million tokens; Highspeed at $4 / $20 (pricing page)
On the internal Kimi Hallu Bench Hard stress test, claim-level hallucination fell from 23.4% (K3) to 8.3%; Moonshot notes this is not a daily-use error rate
Cyber-related data was removed from mid-training and RL for the open-weight release: xdev-bench-hard pass@3 drops from 90.5% (unmitigated checkpoint) to 81.0%

Read full story

Linked Models & Agents:

⚡ Kimi K3.1 Pro⚡ Kimi K3.1 Pro Highspeed⚡ Kimi K3 🤖 Kimi Code 🌐 Moonshot AI

#Kimi#Moonshot AI#open weights#MoE#1M context#coding agent#API pricing

Hugging Face Daily Papers@HuggingFace·Collected 18d ago

👀87

🔥96

MLExpert: Open 35B MoE Agentic Model Lifts MLE-Bench Lite Medal Rate from 37.9% to 72.7%, Matching Kimi-K3 and GLM-5.3

Advancing autonomous AI-for-AI (AI4AI) engineering, researchers from Shanghai AI Lab (InternScience) release MLExpert (arXiv:2610.22344), a 35B Mixture-of-Experts agentic model for autonomous machine learning engineering. Powered by the Kaggle-MLE Factory, hierarchical agentic SFT, and Windowed Interaction RL (WIRL), MLExpert boosts average medal rates on MLE-Bench Lite from 37.9% to 72.7%—matching frontier giants like Kimi-K3 and GLM-5.3 while generalizing to agent scaffold optimization.

Key Takeaways

72.7% medal rate on MLE-Bench Lite: boosts base model performance from 37.9% to 72.7%, matching frontier models Kimi-K3 and GLM-5.3 with a 35B MoE backbone
Windowed Interaction RL (WIRL): assigns shared outcome rewards to multi-turn windows between successive experiment results with asynchronous policy updates
Cross-domain AI4AI generalization: transfers optimization skills directly to NatureBench Lite, agent scaffold optimization, and algorithm engineering

Read full story

Linked Models & Agents:

⚡ MLExpert-35B-MoE⚡ Kimi K3 ⚡ GLM-5.3 🤖 MLExpert 🤖 AutoML Agents 🌐 Shanghai AI Lab 🌐 InternScience

#MLExpert#Machine Learning Engineering#WIRL#MLE-Bench#Kimi-K3#GLM-5.3#AI4AI

Microsoft / Hugging Face@Microsoft·Collected 30d ago

👀129

🔥96

Microsoft×Hugging Face open ThinkingBox: 507 stateful workflows graded on backend state, not chat endings

On 2026-10-03, Microsoft Copilot Studio and Hugging Face released ThinkingBox plus ThinkingBox-Bench (507 tasks × 20 trials): isolated MCP sessions graded by executable terminal-state/side-effect checks, wired into OpenEnv. Vendor board: Claude Opus 5.5 67.16% pass@1; Kimi-K3 93.89% pass@20 but only 13.41% all-20 — one success ≠ reliability. MIT framework microsoft/thinkingbox; CDLA data microsoft/thinkingbox-data; paper arXiv:2608.19741.

Key Takeaways

Sources: HF blog 2026-10-03, OpenEnv docs, microsoft/thinkingbox (MIT), thinkingbox-data (CDLA), arXiv:2608.19741
507 stateful business workflows × 20 isolated MCP trials each
Grade terminal DB/side effects; ~67% of failing traces still terminate cleanly after a mutating tool
Opus 5.5 67.16% pass@1; Kimi-K3 broadest pass@20 (93.89%) but 13.41% all-20; GPT-5.4 ~$6.80/dependable task
Run: OpenEnv + [email protected] → Typesense + MCP proxy → thinkingbox-eval

Read full story

Linked Models & Agents:

⚡ Claude Opus 5.5 ⚡ Claude Opus 5 ⚡ GPT-5.4 ⚡ GPT-6 Astra ⚡ Kimi K3 🤖 ThinkingBox 🤖 OpenEnv 🤖 MCP agents 🌐 Microsoft 🌐 Hugging Face

#Thinkingbox#Microsoft#Huggingface#Openenv#MCP#Agent eval#Benchmark#Reliability

ggml@ggml_org·Collected 50d ago

👀69

🔥92

View on 𝕏 ↗

llama.cpp v0.4.1 adds Maple/Hy4/Spark2.5, qwen3-coder parser fixes, and ggml 0.24.0

Official @ggml_org shipped llama.cpp v0.4.1: adds Maple 20B-A1B, Tencent Hy 4, and Spark2.5 support; improves Kimi-K3 rollback, JSON schema handling, qwen3-coder complex-type parsing, and MCP/multimodal fixes; updates ggml to v0.24.0. Release notes: ggml-org/llama.cpp v0.4.1.

Key Takeaways

Official llama.cpp v0.4.1 adds Maple 20B-A1B, Tencent Hy 4, and Spark2.5.
Improves qwen3-coder parsing, Kimi-K3 rollback, JSON schema, and multimodal/server stability.
ggml bumps to v0.24.0; notes on ggml-org/llama.cpp v0.4.1.

Read full story

Linked Models & Agents:

#llama.cpp#Ggml#v0.4.1#Qwen3 coder#Kimi-K3#local inference#open weights

vLLM@vllm_project·Collected 52d ago

👀85

🔥92

View on 𝕏 ↗

vLLM 0.29.0: Model Runner V2 default, Mamba prefix cache, speculative-decoding stats

vLLM shipped v0.29.0 with Model Runner V2 as the default runner for all models (594 commits, 277 contributors). Highlights include another Kimi-K3 optimization pass, DeepSeek-V4 shared experts fused into MegaMoE, Mamba prefix caching that keeps internal prefill checkpoints (about 9–25% better TTFT), per-request speculative-decoding acceptance stats over the API, a sharded_rdt P2P backend for RL weight sync, and new model support such as Hy4-preview, Qwen3.8-Flash-Next, GraniteSWA, and NemotronH Omni Reasoning V3.

Key Takeaways

Model Runner V2 is now the default for all models.
Mamba prefix caching improves TTFT ~9–25%; speculative decoding exposes acceptance stats via API.
DeepSeek-V4 MegaMoE fusion, RL sharded_rdt P2P sync, and several new model backends.

Read full story

Linked Models & Agents:

⚡ DeepSeek-V4 (Next-Gen MoE)⚡ Kimi K3⚡ Qwen3 8 Flash Next🌐 vllm

#vLLM#Model Runner V2#Mamba#speculative decoding#DeepSeek-V4#Kimi-K3

vLLM@vllm_project·Collected 65d ago

👀48

🔥88

View on 𝕏 ↗

vLLM 0.28.0 Released: Kimi-K3 Full-Stack Optimizations, DeepSeek-V4 Sparse MLA, and 65% TTFT Boost

vLLM 0.28.0 landed with 584 commits, delivering stack-wide Kimi-K3 optimizations, DeepSeek-V4 sparse MLA support, adaptive speculative decoding with 65% TTFT gains, and Model Runner V2 disaggregation.

Key Takeaways

Adaptive speculative decoding with DFlash2/DSpark cuts time-to-first-token by 55-65% across deployments;
Native full-stack acceleration for cutting-edge architectures including Kimi-K3 and DeepSeek-V4 sparse MLA;
Broad hardware enablement across NVIDIA SM12x, AMD ROCm fused kernels, and Intel XPU backends.

Read full story

Linked Models & Agents:

⚡ Kimi K3 ⚡ DeepSeek-V4-Pro

vLLM@vllm_project·Collected 65d ago

👀56

🔥95

View on 𝕏 ↗

vLLM 0.28.0 Released: Full-Stack Kimi-K3, DeepSeek-V4 Sparse MLA, and Tiered Disk KV Cache

vLLM 0.28.0 landed with 584 commits, bringing stack-wide Kimi-K3 support, end-to-end DeepSeek-V4 sparse MLA with DFlash2/DSpark speculative decoding, Model Runner V2 disaggregation, and tiered disk KV cache.

Key Takeaways

Adaptive speculative token budgets slash end-to-end TTFT by 55–65% with 1.5–3x faster sequence-parallel kernels;
End-to-end DeepSeek-V4 sparse MLA and Kimi-K3 shared-expert sharding save ~17 GiB VRAM per GPU;
Tiered KV cache introduces a disk tier, doubles max batched tokens to 16,384, and supports diverse hardware backends.

Read full story

Linked Models & Agents:

⚡ DeepSeek-V4-Flash ⚡ DeepSeek-V3