Writing, research, coding, and files & app tasks — weighed equally.
Highest individual plan. No extra usage purchases.
REPORT FROM THE LOOKOUT
TODAY’S PICKCLAUDE
The latest Opus 5.5 tests give me the best reason to put the needle here, especially for professional work. Its coding results are competitive too. I’d still reach for ChatGPT or Gemini for some research jobs, and writing comes down partly to taste. Claude is my overall pick, not my answer to every question.
This pick has stronger comparative evidence behind it. It can still be wrong for your particular job.
The models, the limits, and the reasons I haven’t moved the needle somewhere else.
GPT-6 Astra · Work · Codex
Astra is close. I’d especially consider it for difficult research or work across connected apps.
Highest individual Pro allowance. Includes the broader Work/Codex ecosystem; extra usage purchases are excluded.
Opus 5.5 · Fable · Cowork / Code
The latest independent work tests are the main reason I’ve picked Opus 5.5.
Highest individual Max allowance. Fable has its own allocation; Opus is the basis of the newest benchmark evidence.
3.8 Flash · Deep Think · Spark
Lots of Google and research tools here. I’d like a current test of all four apps before calling it the winner.
The highest Ultra option. Advanced features have country and rollout restrictions; private model previews are excluded.
Grok ecosystem · 4.7 in Build
4.7 does better at knowledge work and coding. Exactly what Heavy includes is still something I need to pin down.
Heavy is documented, but precise model and effort entitlements remain partially unverified. No private preview is counted.
Here’s what this call rests on. A company talking about its own model is marked “Official.”
Opus 5.5 leads the Intelligence Index and professional-work evaluations; terminal coding is level with Astra. These are model-and-harness tests, not a test of every consumer plan.
Grok improves in professional work and coding. Its latest coding-agent results trail the leading Claude and Astra systems; higher token use complicates efficiency claims.
The latest Opus release adds work and communication improvements. Vendor scores and early-access testimonials are treated as launch evidence, not independent replication.
Google confirms 3.8 Flash for AI Pro and Ultra. The free and AI Plus plans are not assumed to include the same flagship access.
Go is the first paid tier; Pro includes Astra reasoning. Free has limited research and desktop Work access. Live rollout notes take precedence over older model rows.
Pro includes Opus, Research and Claude Code. Max 20x increases capacity. Free includes Sonnet and Haiku, but not Opus or the dedicated Research feature.
Free lists 3.6 Flash, limited 3.1 Pro and Deep Research. AI Plus expands usage; Ultra’s largest option adds more capacity and access to advanced features.
Paid Grok products share a weekly usage pool. The public FAQ describes SuperGrok and Heavy, but does not fully verify the latest model allocation for each plan.
The indexed post announces the new Intelligence Index leader. Full post access was unavailable; the linked evaluation is the substantive evidence, not a second independent vote.
A first-hand account finds both useful and highlights workflow and usage-limit tradeoffs. It predates Opus 5.5, so it provides context rather than a current winner.
The September 23 rollout adds productivity and creative connections, including Airtable, Linear, Adobe and Webflow. The announcement does not specify identical access for every plan or independently demonstrate task reliability.
I wanted a sign that would tell me which AI to open. So here it is. Pick your plan and what you’re trying to do; the needle is the recommendation.
I check which models you can actually get, then look through independent tests, papers, articles, and whatever useful first-hand posts turn up. Writing, research, coding, and files & app tasks count equally in the overall pick. That last category is about producing files and taking actions across apps, rather than just answering a question.
There’s no secret score. If the comparisons are thin, I make a call and mark it as a lean. A missing test doesn’t make a model bad. And a lab announcing something doesn’t mean it’s in your plan yet.
The colored wedges are just the four tools. They aren’t danger levels. This is an independent project, with no connection to any parks department.