Anthropic accelerates Claude Opus 4.6 with a 2.5x fast mode and affirms that Claude will remain ad-free, while OpenAI does exactly the opposite by testing ads in ChatGPT. ElevenLabs triples its valuation to $11 billion, GPT-5.3-Codex lands in GitHub Copilot, and Apple integrates the Claude Agent SDK into Xcode.
Claude Opus 4.6 Fast Mode: 2.5x faster
February 7 — Anthropic launches an experimental “fast” mode for Claude Opus 4.6, offering a token generation speed up to 2.5x faster while maintaining the same intelligence level as the standard model.
The fast mode is available via Claude Code (/fast command), the Anthropic API, and GitHub Copilot (Pro+ and Enterprise in research preview). Boris Cherny, Head of Claude Code at Anthropic, confirms that the internal team has been using it for several weeks and considers it a major productivity unlock for rapid iterations on complex problems.
| Feature | Value |
|---|---|
| Speed gain | 2.5x faster (output tokens) |
| Intelligence | Identical to standard Opus 4.6 |
| Availability | Claude Code (/fast), API, GitHub Copilot (Pro+/Enterprise) |
| Free credits | $50 for Pro and Max |
| Cost | Higher than standard Opus 4.6 |
The mode consumes more compute than standard Opus 4.6 and is therefore more expensive. Anthropic positions it as particularly useful for incident response and projects requiring rapid iteration. On the GitHub Copilot side, a promotional rate of 9 premium requests is offered until February 16.
We just launched an experimental new fast mode for Opus 4.6. The team has been building with it for the last few weeks. It’s been a huge unlock for me personally, especially when going back and forth with Claude on a tricky problem. — @bcherny on X
🔗 Boris Cherny Thread 🔗 GitHub Changelog — Fast Mode Copilot
OpenAI tests ads in ChatGPT
February 9 — OpenAI launches an ad test in ChatGPT with a subset of Free and Go users in the US. Plus, Pro, Business, Enterprise, and Education subscriptions are not affected.
Ads are labeled “Sponsored” and visually separated from responses. OpenAI insists that ads do not influence ChatGPT’s responses and that advertisers have access to neither conversations, history, nor personal data. Users can hide, report, or delete their ad data.
| Principle | Detail |
|---|---|
| Response independence | Ads do not influence responses |
| Labeling | Ads marked “Sponsored”, visually separated |
| Privacy | No advertiser access to conversations or personal data |
| User control | Hide, report, delete ad data |
| Ad-free option | Upgrade to Plus/Pro, or opt-out on Free (fewer daily messages) |
Targeting is done by matching conversation topics, chat history, and past interactions with ads. Advertisers only receive aggregated data. Restrictions in place: no ads for under-18s, no ads near sensitive topics (health, politics), protections against overly narrow targeting and misleading ads.
Anthropic: “Claude will remain ad-free”
February 4 — Just days before OpenAI’s announcement, Anthropic publishes a blog post stating that Claude will remain entirely ad-free. The company explains that ad incentives are incompatible with a truly helpful AI assistant: AI conversations differ fundamentally from search results, users share sensitive information and deep context.
Rather than pursuing an ad-funded model, Anthropic plans to expand access via educational programs in over 60 countries, national AI education pilots with several governments, discounts for nonprofits, and potentially future lower-cost subscription tiers or regional pricing. Anthropic also plans to support user-initiated commercial interactions (agentic purchases, product comparisons) where the AI works exclusively for the user, not advertisers.
ElevenLabs raises 11 billion valuation
February 4 — ElevenLabs announces a 11 billion — more than triple in one year. Total funding reaches $781 million since its founding in 2022.
The round is led by Sequoia Capital, with Andrew Reed joining the board. Andreessen Horowitz (a16z) quadruples its stake, ICONIQ triples theirs, and new investors join: Lightspeed Venture Partners, Evantic Capital, and BOND.
| Metric | Value |
|---|---|
| Amount raised | $500M (Series D) |
| Valuation | $11B |
| Total funding | $781M (5 rounds) |
| ARR end 2025 | > $330M |
| Lead investor | Sequoia Capital |
ElevenLabs closed 2025 with over $330 million in ARR, driven by enterprise adoption with clients like Deutsche Telekom, Square, the Ukrainian government, and Revolut. The company is doubling down on ElevenAgents, its enterprise platform for voice and conversational AI, and announces improvements to its voice agents with faster response times thanks to the Eleven v3 Conversational model.
The funding will strengthen research on emotional conversational models, dubbing, and general audio intelligence. ElevenLabs continues its international expansion with teams in 14 cities (London, New York, San Francisco, Warsaw, Dublin, Tokyo, Seoul, Singapore, Bengaluru, Sydney, São Paulo, Berlin, Paris, Mexico City).
GPT-5.3-Codex GA in GitHub Copilot
February 9 — GPT-5.3-Codex, OpenAI’s latest agentic coding model, is now generally available in GitHub Copilot for Pro, Pro+, Business, and Enterprise users.
| Aspect | Detail |
|---|---|
| Model | GPT-5.3-Codex |
| Tiers | Pro, Pro+, Business, Enterprise |
| Platforms | VS Code, github.com, Mobile, CLI, Coding Agent |
| Performance | +25% vs GPT-5.2-Codex on agentic tasks |
| Rollout | Progressive |
The model sets new record scores on agentic coding benchmarks and improves reasoning in complex long-running workflows. Enterprise and Business admins must enable the GPT-5.3-Codex policy in Copilot settings.
Claude Opus 4.6 free for nonprofits
February 9 — Nonprofits on Team and Enterprise plans now have access to Claude Opus 4.6 at no additional cost. The Claude for Nonprofits program, in partnership with GivingTuesday, offers up to 75% discount.
| Element | Detail |
|---|---|
| Discount | Up to 75% on Team and Enterprise |
| Models included | Opus 4.6, Sonnet 4.5, Haiku 4.5 |
| Open-source connectors | Benevity, Blackbaud, Candid |
| Free course | ”AI Fluency for Nonprofits” |
Three open-source connectors are available to integrate Claude with Benevity (2.4 million validated nonprofits), Blackbaud (CRM and fundraising), and Candid (funder data).
Apple Xcode 26.3 integrates Claude Agent SDK
February 3 — Apple Xcode 26.3 integrates the Claude Agent SDK, allowing developers to leverage autonomous AI capabilities directly within the IDE. This is a significant evolution from the September 2025 announcement which only offered access to Claude Sonnet 4 in Xcode 26.
| Capability | Description |
|---|---|
| Visual verification | Capture and analysis of Xcode Previews |
| Multi-project reasoning | Exploration and understanding of entire projects |
| Autonomous execution | Task decomposition and file modification |
| MCP interface | Claude Code integration via command line |
Xcode 26.3 is available as a release candidate for Apple Developer Program members.
🔗 Apple’s Xcode now supports the Claude Agent SDK
ElevenLabs launches Audiobooks in ElevenCreative
February 9 — ElevenLabs launches Audiobooks, a new tool integrated into ElevenCreative. It is a complete toolkit for creating, refining, and publishing audiobooks with realistic AI voices, from first draft to final publication. The tool targets independent authors, audio content creators, and publishers.
Mistral Voxtral Transcribe 2
February 4 — Mistral launches Voxtral Transcribe 2, two speech-to-text models covering different needs.
| Model | Parameters | Capability | Latency | Price |
|---|---|---|---|---|
| Voxtral Mini Transcribe V2 | — | Batch diarization, 13 languages | — | $0.003/min |
| Voxtral Realtime | 4B | Real-time streaming, open-weights Apache 2.0 | < 200ms | $0.006/min |
Voxtral Realtime is notable: 4 billion parameters, open-weights under Apache 2.0 license, with latency under 200 milliseconds. Mistral claims performance superior to GPT-4o mini and Gemini 2.5 Flash on transcription benchmarks, and 3x higher throughput than ElevenLabs Scribe v2 at one-fifth the price.
xAI expands Grok Imagine with image models
February 8 — xAI extends its Grok Imagine API with image generation and editing models. Initially launched in late January for video and audio, the API now covers the entire creative chain.
| Model | Capability | Arena Benchmark |
|---|---|---|
| Grok Imagine Image Pro | Generation + editing, 10 images/request | Pareto-optimal (single-image-edit) |
| Grok Imagine Image | Text-to-image generation | #4 (score 1170) |
Pricing is positioned at around 4.20/min for video with audio. On February 6, xAI also launched Grok Imagine 1.0 for the general public with 10-second video generation in 720p and improved audio.
What this means
The week illustrates a striking contrast in conversational AI monetization: OpenAI takes the leap into advertising within ChatGPT, while Anthropic publicly takes a stand against it, just days before. Two diametrically opposed visions of the consumer AI business model.
On the technical side, the performance race continues. The Opus 4.6 fast mode and the arrival of GPT-5.3-Codex in Copilot show that inference speed is becoming a major differentiator, not just model intelligence. The integration of the Claude Agent SDK into Xcode confirms Anthropic’s expansion into the Apple ecosystem, a strategic move to reach iOS/macOS developers.
ElevenLabs confirms its trajectory with a tripled valuation and ARR exceeding $330M — a strong signal for the AI audio market. Mistral pushes on open-source speech-to-text with Voxtral Realtime, and xAI completes its creative offering with image generation.
Sources
- Boris Cherny Thread — Fast Mode
- GitHub Changelog — Fast Mode Copilot
- OpenAI — Testing ads in ChatGPT
- Anthropic — Claude is a space to think
- ElevenLabs — Series D
- GitHub Changelog — GPT-5.3-Codex GA
- Anthropic — Claude for Nonprofits
- Anthropic — Apple’s Xcode now supports the Claude Agent SDK
- ElevenLabs — Audiobooks
- Mistral — Voxtral Transcribe 2
- xAI — Grok Imagine API