Weekly models review: 2026-07-20

[github-actions]

Weekly models review (2026-07-20)

Automated check from scripts/weekly-models-check.mjs. Triage and either:

• Update data/pricing.json if a new flagship dropped, a model was deprecated, or pricing changed, then close this issue, OR
• Comment noop and close if nothing actionable surfaced.

Current state of data/pricing.json

• lastUpdated: 2026-07-12 (8 days ago)

• Providers tracked: 13

• Total models: 35

• Flagship models: 16

• Models released within last 90 days: 14

  • 2026-06 | Anthropic | Claude Fable 5 | $10/$50 | flagship
  • 2026-06 | Anthropic | Claude Sonnet 5 | $2/$10 | mid
  • 2026-05 | Anthropic | Claude Opus 4.8 | $5/$25 | flagship
  • 2026-06 | OpenAI | GPT-5.6 Sol | $5/$30 | flagship
  • 2026-06 | OpenAI | GPT-5.6 Terra | $2.5/$15 | mid
  • 2026-06 | OpenAI | GPT-5.6 Luna | $1/$6 | budget
  • 2026-05 | Google | Gemini 3.1 Flash-Lite | $0.25/$1.5 | budget
  • 2026-05 | Google | Gemini 3.5 Flash | $1.5/$9 | mid
  • 2026-07 | Meta | Muse Spark 1.1 | $1.25/$4.25 | mid
  • 2026-05 | Alibaba | Qwen3.7-Max | $2.5/$7.5 | flagship

Model-release-flavored news, last 7 days

Matched 12 articles (keyword scan; not all will be real releases).

Date	Source	Title
2026-07-20	The Verge AI	China delivers a one-two punch to America’s AI dominance
2026-07-18	WIRED AI	How Google’s New Gemini Rates Work and How to Track Your Usage
2026-07-18	WIRED AI	Prompt Injection Attacks Are Thwarting AI Hacking Agents
2026-07-16	Google AI Blog	Create, edit and star in videos with two Google Vids updates
2026-07-16	The Verge AI	Google is renaming NotebookLM to Gemini Notebook
2026-07-16	The Verge AI	Claude can now use your 1Password credentials for you
2026-07-16	MIT Technology Review	The Download: OpenAI unveils GPT-Red and heat pumps rise in the US
2026-07-15	NVIDIA AI Blog	NVIDIA Introduces New Jetson Thor Computers to Advance Mainstream Robotics and Edge AI
2026-07-15	VentureBeat AI	Agentic orchestration: Enterprise AI organizations have a deployment problem, not a platform problem — and most are calling chatbots agents
2026-07-15	MIT Technology Review	Meet GPT-Red: an LLM super-hacker OpenAI built to make its models safer
2026-07-15	Hugging Face Blog	Introducing Real World VoiceEQ: Measuring the human quality of voice AI
2026-07-14	NVIDIA AI Blog	Nemotron Labs: How Open Models Give Enterprises and Nations AI They Can Trust, Control and Customize

Sources: The Verge AI (3), WIRED AI (2), MIT Technology Review (2), NVIDIA AI Blog (2), Google AI Blog (1), VentureBeat AI (1), Hugging Face Blog (1)

Provider model counts

Provider	Model count
Anthropic	7
OpenAI	8
Google	4
Meta	3
Mistral	2
Cohere	2
DeepSeek	2
Alibaba	1
xAI	2
NVIDIA	1
Microsoft	1
MiniMax	1
Meituan	1

---

What "needs update" usually means

1. A new flagship model shipped from a tracked provider. Add it to that provider's models[].
2. A tracked model was deprecated or sunset. Either remove the entry or mark its tier appropriately.
3. A provider changed pricing for a tracked model. Update inputPrice / outputPrice.
4. A new provider has become noteworthy enough to add. Add a new providers[] entry.

What it usually does NOT mean

• Research papers about models (those land on /research).
• Beta or preview models that are not generally available.
• Open-source model fine-tunes (those belong in /best-open-source-llms if anywhere).

Bump lastUpdated in data/pricing.json whenever you change anything else in the file. Also bump last_reviewed for /models in data/page-freshness.json.