
Anthropic Lost $8B Operating in 2025. You Feel That as Usage Caps.
Reuters: Anthropic 2025 operating loss $8.06B on $4.6B revenue, $42B net with a $34B paper charge. OpenAI halved $200 Pro usage. Consumers pay now.

Expert profile
Founder & AI Product Leader
Yash Thakker is a Generative AI expert with over 12 years of experience in product leadership and technical strategy. As the founder of explainx.ai, he has taught over 300,000 learners and built AI platforms serving millions of users globally. He specializes in Agentic AI, Multimodal RAG, and the intersection of LLMs with consumer hardware.

Reuters: Anthropic 2025 operating loss $8.06B on $4.6B revenue, $42B net with a $34B paper charge. OpenAI halved $200 Pro usage. Consumers pay now.

Baselayer's Agentic Identity Suite adds Know Your Agent (KYA) so banks can verify who an agent represents before it transacts — vs OpenShell and wallets.

OpenAI’s Pro models table shows Unlimited* on GPT-5.6 Sol, Luna, and Terra. GPT-6 stays Expanded. Asterisk is fair use, not Codex hours.

Some AI and chip exec families need Beijing travel approval, Bloomberg relays say. What hiring and M&A teams should track.

Anthropic’s Lance Martin added /claude-api build-eval and hillclimb to Claude Code on Sep 28, 2026 — production-shaped evals, held-out tests, 90.5% vs 78.6%.

Opus 5.5 scores 58 on AA vs Sonnet 5.5’s 56. List price is 2×; at max effort Sonnet can cost more per task. When to keep the Claude Code default.

Sonnet 5.5 launched Sep 28, 2026: 30% faster, up to 30% cheaper per task. Model ID claude-sonnet-5-5, plus Osmani’s API and Claude Code migration.

Sonnet 5.5 scores 56 on Artificial Analysis vs GPT-6 Astra’s 53 at max — at $2/$10 vs $10/$50. Token burn and Terminal-Bench decide the real pick.

ElevenLabs launched Eleven v4 and v4 Turbo on Sep 28, 2026 — inline audio tags, ~100ms Turbo latency, 90+ languages, PVC support. Artificial Analysis #1 rank.

Florida AG James Uthmeier asked a judge to bar OpenAI from new models without oversight, keep minors off ChatGPT, and end anthropomorphic marketing (Reuters).

GPT Researcher 3.7 uses Jev as an optional usefulness filter. No TypeSafe key means BM25, not embeddings. 73% is kept-passage precision on 28 tasks.

Grok 4.7 landed on Amazon Bedrock Sept 28: 500K context, four effort levels, Converse plus OpenAI APIs. When AWS shops should leave the xAI key.