Isomorphic Labs raises $5 billion in Series C

Greenoaks and Sequoia Capital co-led the round, with further investment from Alphabet. Isomorphic plans to expand its IsoDDE drug design models and bioengineering infrastructure and advance candidates toward clinical trials. The oncology and immunology progress cited in the announcement is preclinical.
Genesis AI raises $500 million in Series A
Premji Invest, Khosla Ventures and Eclipse co-led the round at a reported $3.5 billion post-money valuation. Genesis will use the funding to expand data collection, training and deployment of its flagship robot, Eno.
NVIDIA reaffirms plans to acquire Hugging Face
TechCrunch reported that interim CEO Colette Kress reaffirmed NVIDIA’s open AI strategy and acquisition plans. The companies signed a definitive agreement in September, and the deal is still pending.
Jason Wei joins Mercor as chief scientist
Jason Wei announced his move to Mercor after a year at Meta’s TBD lab. He singled out training data as a focus for the next stage of model development.
Three Meta Muse models are coming to Microsoft Foundry
Meta confirmed that Muse Wonder, Muse Spark 1.3 and Muse Ember will be offered on Azure through Microsoft Foundry. The announcement gave no launch date.
IFM publishes the K2 Horizon technical report
The report documents six previously announced models, from a 0.9-billion-parameter dense model to a 375-billion-parameter MoE with 23 billion active parameters. The project provides weights, training data, code and intermediate checkpoints.

Hugging Face releases a Qwen3.8‑27B adapter for Pi
The team trained a LoRA adapter with reinforcement learning for the Pi coding agent. In the team’s pass@1 tests using the same harness, SWE‑bench Pro Hard 51 rose from 29.4% to 62.7%, and Terminal‑Bench 2.1 rose from 39.5% to 55.8%.
Gemini Tasks expands free access to six countries

The Gemini 4 Hydrogen update is rolling out free in the US, Japan, South Korea, Canada, Australia and India. Some advanced features start in the US, and critical actions like purchases prompt for confirmation or pre-approval.
Replit Missions gives agent teams a budget

Users define the outcome, spending cap and actions that require approval. Agents then plan, execute and test in parallel. In its own example, Replit used a $3,000 budget and two days to build an internal application monitoring tool.

Antigravity plugin connects robot arms and MuJoCo
Google DeepMind demonstrated natural-language control of a supported robot arm with Gemini 4 Hydrogen and the Antigravity robotics plugin. Users can first try the task in MuJoCo and then run it on physical hardware.
Claude Code experiments with chaining tools in scripts
Code Mode lets Claude write scripts that connect multiple tools and keep responding while long commands are still running. Set CLAUDE_CODE_EXPERIMENTAL_CODE_MODE=1 to opt in.
Claude Platform tests linked billing

Customers of tools such as Factory, Linear and Ramp can charge Claude API usage to their own Claude Platform organization without giving the integration an API key.
OpenAI updates its API usage dashboard

Analytics lets teams track usage and spending, inspect rate limits and rate-limited requests, and investigate errors by model, project or API key.
Perplexity introduces the Orchestrator API

The API brings open-weight and proprietary models under one interface and bill. Callers can choose a model, while the service selects providers based on availability, capacity and latency and supports failover.
Perplexity rolls out Multiagent

Computer and Teammates are getting a Multiagent effort setting. A lead agent delegates work, agents using different models trade findings, and the lead assembles the answer.
Replit Desktop gives Agent access to selected folders

Users choose which folders the Agent can use to build and run apps. On supported Windows and macOS devices, Microsoft Execution Containers provide a local sandbox.
Replit open-sources DataCat for self-hosted observability
DataCat handles metrics, logs, traces, dashboards and alerts, and accepts OpenTelemetry data. It supports PostgreSQL and ClickHouse and can be self-hosted from the public repository.
Vercel releases eve 1.0
The open-source framework is generally available and targets agents that keep running across multiple steps. Developers can integrate and deploy their own agents in applications.
OpenHands ships a lighter enterprise edition
The OpenHands team highlighted using cloud coding agents for security audits and remediation, software testing, and code review.
Claude Desktop adds a Latest release channel
Windows and macOS users can opt into Latest in Desktop app settings for earlier fixes. Stable remains the default, and Latest builds can be less reliable.
Adaption adds a Hermes Agent plugin
Adaption announced a plugin for preparing data and training custom models from inside Hermes Agent.

Apple releases FastFrames video generation models
The model cards list FastFrames‑S at 1.3 billion parameters and FastFrames‑L at 14 billion. Both use two-step autoregressive video diffusion. The published configuration generates 832×480 video at 16 fps for 81 frames, with interactive latency as the project’s goal.
SGLang summit recap details version 1.0 and nine founding members

The SGLang team recapped the announcement of SGLang 1.0, Miles 0.2 for reinforcement learning and the SGLang Consortium. Its founders include NVIDIA, AMD, Intel, Google, Meta, Thinking Machines, Cloudflare, Modal and RadixArk.
vLLM updates DSpark for Kimi K3.1 Pro

Inferact and Red Hat collaborated on the revised speculative decoding draft model. vLLM reports up to 30% longer accepted drafts than the preview on HumanEval and a cap on the draft model’s KV cache.
vLLM‑Omni adds day-one support for AURA 2.0

The 9-billion-parameter video model follows a scene as it unfolds, can initiate a conversation and can call tools in response to what it sees.
DeepsecBench adds GPT‑6.2 vulnerability detection results

Vercel added GPT‑6.2 to its application security leaderboard. Under Codex, Astra at xhigh reached 56.2% recall and 95.0% precision, with a reported test cost of $59.85.
Perplexity reports per-task costs for both GPT‑6.2 models
Both models are in Search and Computer. On Perplexity’s WANDR benchmark, Astra scored 0.783 at about $12.54 per task and Sol scored 0.746 at about $2.65 per task.
Calif uses GPT‑6 Cyber to help find a Java flaw
Greg Brockman said security team Calif used GPT‑6 Cyber to help discover a major Java security flaw under the Patch the Planet initiative.
Ollama outlines cache pitfalls in long coding sessions
Ollama shared observations from Claude Code, Codex and Pi. Timestamps in system prompts and changes to reasoning effort can invalidate the cache. Telling the agent that output was truncated can prompt it to read what it missed.
Inferact revisits pipeline parallelism for prefill
The team argues that long-running agent tasks may change prefill traffic patterns. Idle time from pipeline bubbles is worth measuring again under those workloads.
NVIDIA releases a speculative decoding draft for Qwen3.8‑27B
The 4-billion-parameter draft model proposes candidate tokens in batches for the full Qwen3.8‑27B model to verify.
vLLM supports Sage S2.1 Flash
The model combines scientific inputs such as molecular graphs and protein structures with retrieval, coding and tool calls. It can check intermediate results and refine its next step.
An energy function gives recurrent reasoning a stopping signal

Energy‑Based Reasoners learn an energy function over latent reasoning states. Gradient descent refines those states and helps decide when to stop. The paper compares the method with earlier recurrent models on four algorithmic reasoning benchmarks.
Ropedia Xperience 2 aligns scenes with human motion

The paper releases 1,589 metric-scale 3D scans and 4,230 first-person recordings covering 74 tasks over 131 hours. Hand and full-body motion is aligned to the scenes for simulation and robot trajectory retargeting.
Accelerate Bench tests whether agents can speed up programs

The benchmark includes 32 CPU program optimization tasks, with independent verification of correctness and measured speedup. Across 16 model-and-harness configurations, the authors report Score@1 results from 25.5% to 67.4%.
V‑JEPA‑LM trains on counterfactual video to read motion
The work pairs a frozen V‑JEPA 2.1 encoder with language models. The experiments cover models from 1 billion to 8 billion parameters and use counterfactual video to train for motion understanding.
Simon Willison demonstrates external sandboxes
The coding agent stays on the Mac, while its generated code executes in WebAssembly, QEMU or Apple Container. The demonstrations cover Claude Code, Codex and Pi.
Raschka discusses constant learning rates and weight averaging
Raschka noted that some pretraining recipes keep the learning rate constant and average late checkpoints or maintain a running weight average. That removes the need to schedule learning-rate decay ahead of time.
Mollick reports uneven results with Astra

Ethan Mollick found GPT‑6.2 Astra harder to steer in early use. In the same trials, it could tackle a demanding prompt to build a space elevator scene from scratch using Blender and Unity.
One prompt, two GPT‑6.2 takes on a Three.js scene

Simon Willison gave Astra and Sol the same prompt: use OpenStreetMap data to turn Yonkers into a mech that walks through several New York City boroughs. He shared both outputs.
Runway Labs shows Diffusion‑UI
In a research demo shared by Runway, a real-time model generates the pixels of an interactive interface. The project explores how image generation might become part of the interface itself.
Midjourney’s edit model gains style references

After attaching a base image in the prompt bar, users can add a Moodboard, Style Code or Style Reference and submit the edit. Midjourney’s example needs no text prompt.
Ideogram opens its Print on Demand Studio

The workspace includes research, product templates, design, prepress and previews of finished products. Ideogram’s Vector model is available for image generation.
Kling demonstrates native 8K with Image 4.0

Following its Image 4.0 preview, Kling shared a close-up demonstration highlighting native 8K image generation.
Vidu launches MCP and a CLI

The new interfaces let MCP-enabled agents and terminal users call Vidu directly to launch video generation from a prompt.
PixVerse Canvas adds a 3D Director Console

Creators can position actors, place the camera and frame the scene before generating the shot.
