🚀 Anthropic launches Claude Opus 4.7 with 3x task resolution 🚀
Opus 4.7 raises Anthropic's coding ceiling and triples image resolution across all products.

Anthropic released Claude Opus 4.7, a direct upgrade to Opus 4.6 that delivers state-of-the-art results in software engineering, with a 13% lift on CursorBench and 3× more production tasks resolved on Rakuten-SWE-Bench. The model accepts images up to 3.75 megapixels — over three times prior Claude models — unlocking dense screenshot reading and pixel-precise document extraction. A new xhigh effort level gives developers finer control over reasoning depth versus latency. Pricing stays at $5/$25 per million input/output tokens, matching Opus 4.6, which keeps the upgrade cost-neutral for existing users. Opus 4.7 ships with built-in cybersecurity safeguards ahead of Anthropic's eventual broad release of its more powerful Mythos-class models, a clear signal of tighter capability gating at the frontier.

🔊 Google ships Gemini 3.1 Flash TTS model 🔊
Google's new TTS model brings granular voice control to over 70 languages.

Google released Gemini 3.1 Flash TTS, a text-to-speech model that scored 1,211 Elo on the Artificial Analysis TTS leaderboard, landing at #2 overall. The model introduces audio tags that let users control vocal style, tone, pace, and accent through natural language commands — a notable capability jump for customizable speech. All generated audio is watermarked with Google's SynthID technology to flag synthetic content. Supporting over 70 languages, the model competes directly with OpenAI and ElevenLabs in a TTS market where competitive pressure is intensifying rapidly. The combination of expressivity, multilingual reach, and built-in safety features makes this one of the most complete TTS rollouts to date.

🎬 NVIDIA rolls out Lyra 2.0 for 3D video 🎬
NVIDIA's framework tackles spatial consistency and temporal drift in long AI videos.

NVIDIA released Lyra 2.0, a framework for generating long, camera-controlled videos that maintain 3D spatial consistency throughout. The system uses geometry-guided retrieval to prevent spatial forgetting and self-augmented training to reduce temporal drift — two persistent problems in video generation. Lyra 2.0 converts text prompts and camera path descriptions into explorable 3D scenes, a capability jump with direct applications in content creation, gaming, and simulation. This positions NVIDIA not just as an AI hardware leader but as a serious competitor in creative AI tools, where speed and coherence over longer sequences have remained stubborn challenges for existing models.

INTERESTING TO KNOW

🛠️ OpenAI adds sandboxed execution to Agents SDK 🛠️

OpenAI updated its Agents SDK with a model-native harness for cross-file and tool workflows, plus sandboxed execution for safer task handling. The sandboxed environment limits what agents can do within controlled boundaries — an important integration layer for preventing unintended actions in production settings. These updates aim to make enterprise-grade agent development more secure, a clear adoption signal as competition from LangChain, Google's ADK, and Anthropic's tool ecosystem intensifies.

👟 Allbirds pivots from sneakers to GPU rentals 👟

Sustainable-footwear company Allbirds announced a $50M financing deal to reinvent itself as "NewBird AI", a GPU rental business — a pricing play on surging AI compute demand that sent its stock up over 600%. The company had sold its brand assets for $39M in March, a steep fall from its $4B IPO valuation in 2021. Shareholders will vote next month to strip the company's "public benefit" status, formally ending its sustainability mission — a stark signal of how intense competitive pressure for compute infrastructure has become.