Frontier models arrive everywhere at once; the work is in the wiring
GPT-5.4, ChatGPT for Excel and Gemini 3.1 Flash-Lite, and The Deploy Log for this week. Every price and claim read off the publisher's own page, with a call on each.
OpenAI
GPT-5.4
DEPLOYif you do professional knowledge work
WHAT SHIPPEDGPT-5.4 is released in ChatGPT as GPT-5.4 Thinking, the API, and Codex, with GPT-5.4 Pro also available. It scores 83.0% on GDPval, 57.7% on SWE-Bench Pro, 75.0% on OSWorld-Verified, and supports up to 1M tokens of context.
THE CATCHThe model is the first general-purpose release with native computer-use capabilities, and it is the most token efficient reasoning model yet, using significantly fewer tokens than GPT-5.2.
IF YOU RUNa team doing spreadsheets, presentations, or documents, GPT-5.4 scores 87.3% on an internal investment banking spreadsheet benchmark versus 68.4% for GPT-5.2
SOURCEOpenAI, Introducing GPT-5.4
OpenAI
ChatGPT for Excel
DEPLOYif your team works in Excel
WHAT SHIPPEDChatGPT for Excel is in beta, an Excel add-in that builds, updates, and analyzes spreadsheet models directly in workbooks. It is available globally to ChatGPT Business, Enterprise, Edu, Teachers, and K-12 users, and to ChatGPT Pro and Plus users outside the EU.
THE CATCHOn OpenAI's internal investment banking benchmark, performance improved from 43.7% with GPT-5 to 87.3% with GPT-5.4 Thinking, and the add-in asks for permission before making changes so users can review each step and undo edits.
IF YOU RUNfinancial modeling, scenario analysis, or budgeting in Excel, ChatGPT for Excel can reason across workbooks, trace and fix errors, and show how assumptions flow through a model
SOURCEOpenAI, Introducing ChatGPT for Excel and new financial data integrations
Google DeepMind
Gemini 3.1 Flash-Lite
DEPLOYif you run high-volume workloads
WHAT SHIPPEDGemini 3.1 Flash-Lite is rolling out in preview via the Gemini API in Google AI Studio and Vertex AI, priced at $0.25 per 1M input tokens and $1.50 per 1M output tokens. It achieves 86.9% on GPQA Diamond and 76.8% on MMMU Pro.
THE CATCHIt outperforms 2.5 Flash with a 2.5X faster Time to First Answer Token and 45% increase in output speed, and it comes standard with thinking levels so developers can control how much the model thinks per task.
IF YOU RUNtranslation, content moderation, or data extraction at scale, 3.1 Flash-Lite is built for high-volume developer workloads where cost and latency are the primary constraints
SOURCEGoogle DeepMind, Gemini 3.1 Flash-Lite: Built for intelligence at scale
The Deploy Log
Models and APIs
| DEPLOYif you use ChatGPT daily GPT-5.3 Instant. GPT-5.3 Instant is released as an update to ChatGPT's most-used model, reducing unnecessary refusals and improving web answer synthesis. OpenAI |
| WATCHif you build custom diffusion workflows Modular Diffusers. Modular Diffusers introduces a new way to build diffusion pipelines by composing reusable blocks, with a quickstart example running FLUX.2 Klein 4B and community pipelines including Krea Realtime Video achieving 11fps on a single B200 GPU. Hugging Face |
Tools and products
| DEPLOYif you review code for vulnerabilities Codex Security. Codex Security rolls out in research preview to ChatGPT Pro, Enterprise, Business, and Edu via Codex web, free for a month; scanned 1.2M+ commits in 30 days, finding 792 critical and 10,561 high-severity issues. OpenAI |
| DEPLOYif you use Notion AI GPT-5.4 in Notion. GPT-5.4 is now available in Notion's model picker, bringing faster responses and better token efficiency for demanding tasks. Notion |
| DEPLOYif you route models through AI Gateway GPT-5.4 on AI Gateway. GPT-5.4 and GPT-5.4 Pro are now available on AI Gateway, bringing agentic and reasoning leaps from GPT-5.3-Codex to all domains including knowledge work and coding. Vercel |
| DEPLOYif you run high-volume agentic tasks Gemini 3.1 Flash Lite. Gemini 3.1 Flash Lite from Google is now available on AI Gateway, outperforming 2.5 Flash Lite on overall quality with notable improvements in translation, data extraction, and code completion. Vercel |
| DEPLOYif you use AI Gateway for chat GPT-5.3 Chat on AI Gateway. GPT-5.3 Chat (GPT 5.3 Instant) is now available on AI Gateway, focusing on tone, relevance, and conversational flow with fewer unnecessary refusals and caveats. Vercel |
| DEPLOYif latency compounds in your agent loops Mercury 2 on AI Gateway. Mercury 2 from Inception is now available on AI Gateway, delivering reasoning-grade quality at real-time latency for agentic loops, coding assistants, voice interfaces, and RAG pipelines. Vercel |
| DEPLOYif you run basic tasks in Notion agents MiniMax M2.5 in Custom Agents. Custom Agents now support MiniMax M2.5, an open weight model that is up to 10x more cost-efficient for basic tasks, available in the Custom Agent model picker. Notion |
Breakthroughs
| DEPLOYif you build speech systems for African languages WAXAL speech dataset. WAXAL is a large-scale, openly accessible speech dataset covering 27 Sub-Saharan African languages spoken by over 100 million speakers, with approximately 1,846 hours of transcribed natural speech for ASR and over 565 hours of high-fidelity recordings for TTS, released under CC-BY-4.0. Google Research |
| DEPLOYif you process camera trap images SpeciesNet wildlife identification. SpeciesNet is an open-source AI model that classifies 2,498 animal categories in camera trap images, trained on over 65 million labeled images, and it finds 99.4% of images containing animals with 94.5% of species-level predictions correct. Google Research |
Left out
- GPT-5.4 Pro: No separate page text provided for GPT-5.4 Pro beyond its mention within the GPT-5.4 release page.
- ChatGPT for Google Sheets: The page text mentions ChatGPT for Google Sheets is coming soon and later generally available, but the primary item is ChatGPT for Excel and
- Financial data integrations in ChatGPT: The page text lists integrations with FactSet, Dow Jones Factiva, LSEG, Daloopa, S&P Global, Moody's, MSCI, Third Bridge, and MT Newswires,
- Codex for OSS: The page text describes onboarding an initial cohort of open-source maintainers into Codex for OSS, but the item is part of the Codex Securi
- Mellon integration with Modular Diffusers: The page text states Mellon is in early development and not ready for production use yet, so it does not meet the deployment lens threshold.
- Krea Realtime Video: The page text describes Krea Realtime Video as a community pipeline built with Modular Diffusers, but it is part of the Modular Diffusers pa
- Waypoint-1: The page text describes Waypoint-1 as a community pipeline built with Modular Diffusers, but it is part of the Modular Diffusers page and do
- SpeciesNet partner projects: The page text describes multiple partner projects using SpeciesNet, but the item is part of the SpeciesNet page and does not have a separate
- WAXAL derivative research: The page text describes derivative research enabled by WAXAL, but the item is part of the WAXAL page and does not have a separate score or l
- NXP i.MX 95 SoC: The page text describes the i.MX 95 SoC as the hardware platform for VLA optimization, but the item is part of the VLA on embedded platforms
- Stripe on Marketplace: a figure in the draft could not be found on the publisher's own page
- v0 API custom MCP servers: the publisher's page does not show it available yet
- VLA on embedded platforms: the publisher's page does not show it available yet
- Provider-level custom timeouts: a figure in the draft could not be found on the publisher's own page
- Vercel Workflow speedup: the publisher's page does not show it available yet
- Can create pages permission: it did not carry a number that moved
- Sandbox environment variables: it did not carry a number that moved
- CDN routing rules: it did not carry a number that moved
- Chat SDK table rendering: it did not carry a number that moved
- MCP Apps on Vercel: it did not carry a number that moved
- Responses API on AI Gateway: it did not carry a number that moved
The pattern
GPT-5.4 arrives in ChatGPT, the API, Codex, Notion, and Vercel AI Gateway in the same week that Gemini 3.1 Flash-Lite and Mercury 2 land on AI Gateway, while ChatGPT for Excel, Codex Security, and GPT-5.3 Instant ship as adjacent products. The frontier model is no longer the scarce asset; the scarce asset is the wiring that puts a model inside a workbook, a gateway, a security scanner, or a custom agent. OpenAI's own investment banking benchmark jumps from 43.7% with GPT-5 to 87.3% with GPT-5.4 Thinking, but that gain only becomes usable when ChatGPT for Excel asks permission before each edit and lets users undo changes.
The same pattern holds at lower cost and higher speed: Gemini 3.1 Flash-Lite at $0.25 per 1M input tokens and $1.50 per 1M output tokens outperforms 2.5 Flash with a 2.5X faster Time to First Answer Token, and MiniMax M2.5 in Notion Custom Agents is up to 10x more cost-efficient for basic tasks. Meanwhile, Modular Diffusers turns diffusion pipelines into reusable blocks running FLUX.2 Klein 4B and Krea Realtime Video at 11fps on a single B200 GPU, and WAXAL releases 1,846 hours of transcribed speech across 27 Sub-Saharan African languages under CC-BY-4.0. The work is not in the model weights; it is in the add-ins, gateways, pipelines, and datasets that decide where a model can act.