- Bay Area Times
- Posts
- GPT-5.4 debuts as OpenAI’s “most capable” frontier model
GPT-5.4 debuts as OpenAI’s “most capable” frontier model
Bay Area Times
March 06, 2026
Brought to you by:
Top stories today:
- GPT-5.4 debuts as OpenAI’s “most capable” frontier model
- Amodei says Anthropic will fight DoW risk label
- Anduril said to see revenue double to ~$4.3B, losses widen to $1.2B
- Anthropic study finds AI mostly augmenting — not replacing — workers
- OKX secures investment from ICE, NYSE parent, at $25B valuation
- Together AI said to seek ~$1B raise at $8.5B post-money valuation
0. Data and calendar
All values as of 6 AM ET / 3 AM PT, other than S&P500 and NASDAQ close (4 PM ET / 1 PM PT).
All times are ET.
Listen to our AI-generated podcast summarizing today’s newsletter (beware of hallucinations):
1. GPT-5.4 debuts as OpenAI’s “most capable” frontier model
GPT-5.4 Pro ranks #2 slightly behind Gemini 3.1 Deep Think on the ARC-AGI-2 leaderboard with 83.3% at $16.41 per task:
GPT-5.4 Thinking is available to ChatGPT Plus, Team, and Pro users, replacing GPT-5.2 Thinking.
GPT-5.4 Pro is available via the API and for ChatGPT Enterprise and Edu users.
1M-token context window on GPT-5.4 Codex, up from 400K on GPT-5.3 Codex.
The model produces better presentations with more varied aesthetics.
GPT-5.4 Thinking lets users adjust course mid-response in ChatGPT.
Improves deep web research for specific queries and maintains context for longer reasoning, OpenAI said.
GPT-5.4 adds computer-use capabilities in Codex and the API.
GPT-5.4 costs $2.5/$15 per 1M input/output tokens, while GPT-5.4 Pro costs $30/$180, vs. GPT-5.2 ($1.75/$14), Claude Opus 4.6 ($5/$25), and Gemini 3.1 Pro ($2/$12 ≤200K tokens).
GPT-5.4-high ties with Gemini 3 Pro on Arena’s user-voted leaderboard but trails Gemini 3.1 Pro, Claude Opus 4.6, and Grok 4.20
ChatGPT for Excel add-in has debuted for paid users in the U.S., Canada, and Australia
Spreadsheet and presentation capabilities in Codex and the API have been updated.
Financial data integrations are coming to ChatGPT from FactSet, Dow Jones Factiva, LSEG, Daloopa, and S&P Global.
Post-training, evaluation, reasoning research — this is PhD-level work, and U.S. supply just isn’t keeping up with demand.
Athyna Intelligence connects you with vetted AI researchers from Latin America who bring deep STEM expertise, move fast, and work in U.S.-aligned time zones. Oh, and at 40–60% less than hiring domestically.
The talent exists. You just have to know where to look.
*Disclaimer: We have equity in Athyna.
3. Amodei says Anthropic will fight DoW risk label, apologizes for leaked memo
“Anthropic did not leak” the memo referenced in reports, he said.
Anthropic and its products were deemed a supply chain risk effective immediately, the DoW said.
Claude downloads were +220% on Tue. vs. Feb. 23, per Apptopia.
Sam Altman took a swipe at Anthropic, saying abandoning democratic commitments based on who is president is “bad for society.”
