AI News

Gemini

Gemini 4 Argon: What's New, Who Can Use It, Pricing, and Benchmarks

A practical guide to Gemini 4 Argon: who can use it in the Gemini app and on which plan, why auto model selection and the default thinking level matter, what Tasks and the new coding interface add, the introductory API price and what it becomes in December, where Google’s own benchmarks show Argon ahead and behind, independent Artificial Analysis scores, safety findings from the model card, and where Google’s own pages disagree.

Meta Muse Wonder vs Wonder Full: What's New, Who Can Use Each, Pricing, and Benchmarks

A practical guide to Meta’s Muse Wonder release: why only the smaller of the two models is open to developers, the API price against Muse Spark and Claude Opus 5.5 and what cost per task does and does not tell you, Standard versus Contributor data terms, what Active Memory and Active Focus mean, where Muse Wonder Full leads and trails in Meta’s benchmarks, independent Artificial Analysis scores, what changes for Muse app users, and what has been reported but not published.

ChatGPT

ChatGPT Dots: What the Always-On Agent Does, Who Gets One, and How Usage Is Counted

A practical guide to OpenAI’s dots, announced at DevDay on September 29, 2026: what an always-on agent with its own cloud computer does between conversations, exactly who is eligible, how action review and custom rules bound its autonomy, the three separate usage limits, what pausing and deleting do and do not undo, and what OpenAI has not yet published.

Claude Sonnet 5.5

Claude Sonnet 5.5: Faster at the Same Token Price, but Not Always Cheaper per Task

Claude Sonnet 5.5 costs the same per token as Sonnet 5 and generates output faster. Whether it costs less per task depends on the effort setting: Artificial Analysis measures a saving at Medium and a higher bill at Max, where Opus 5.5 at Xhigh matches its score for less. This guide covers the benchmark table with Anthropic’s footnotes, the six changes that can break API integrations, the first cyber safeguards on a Sonnet model, and how to divide work between Sonnet 5.5 and Opus 5.5.

ChatGPT

ChatGPT Voice Adds Plugins to Live and Expands Work Access: Where It Works and How Usage Is Counted

OpenAI’s Voice update makes connected tools available during Live conversations and brings Voice in ChatGPT Work to web and mobile. The details depend on where the conversation starts: Chat, Work, and the desktop app have different model access and usage rules. This guide separates the three, lists what is documented for each, and covers approvals, limits, and privacy controls.

GPT-6

GPT-6 Sol and Luna: What's New, Pricing, Benchmarks, and Who Should Use Them

A practical guide to OpenAI’s GPT-6 Sol and Luna: the token price cuts and what they do and do not guarantee, the new five-hour usage limits in ChatGPT Work and Codex, coding and factuality claims, benchmark results with their footnotes, availability by plan, and which model fits which job.

Claude Opus 5.5

Claude Opus 5.5: What's New, Pricing, Benchmarks, and Who Should Use It

A practical guide to Claude Opus 5.5: agentic coding upgrades, the pricing table against Opus 5 and Fable 5.1, how the 40% workload-savings claim is built and what independent testing shows so far, benchmark results with their footnotes, safety changes including the loss of thinking-off mode, availability, and who should switch.

ElevenLabs

ElevenLabs Studio 4.0 Adds In-Editor Generation, but Per-Video Costs Remain Opaque

Studio 4.0 moves media generation inside ElevenLabs’ editor and rebuilds the timeline for multi-scene projects. This piece lists what ElevenLabs announced, which third-party models sit behind the generate button, what each plan includes, and why the cost of a finished video cannot be worked out from published information.

Grok

Grok 4.7 Keeps the $2 and $6 Token Rates, but xHigh Benchmarks and Higher Token Use Complicate the Upgrade

Grok 4.7 ships at Grok 4.6’s starting token rates with higher vendor benchmark scores, but most of SpaceXAI’s comparisons pair Grok 4.7 at xHigh effort with Grok 4.6 at High, and independent testing shows the new model generating far more tokens per task. Here is what is published, what Artificial Analysis measured, and how to run a fair test.

Claude

Claude Is Merging Chat and Cowork and Adding Docs and Slides: Plans, Exports, and Open Questions

Anthropic has started merging Claude chat and Cowork into one interface and added Claude Docs and Claude Slides in beta, with Claude Design now available inside conversations. This explainer covers rollout by plan, what each tool creates and exports, and the details Anthropic has not documented, including PowerPoint editability and real-time co-editing.