What Each Tool Brings To My September 2026 AI Stack

🔍 Read the full analysis: What Each Tool Brings To My September 2026 AI Stack on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on office and shipping supplies

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

Thorsten Meyer published his September 2026 AI stack on 29 September 2026, assigning Claude Opus 5.5 as his main building model and the same-day-launched GPT-6.1 Sol as his routine review model. The review argues that the frontier has compressed into a price curve: six models now sit within roughly 20 index points of each other on the Artificial Analysis Intelligence Index v4.3.x, while their cost per task differs by about 100x — shifting the practical question from “which model is smartest?” to “which model clears my quality bar at the lowest cost per task?”

All capability scores in the review come from the Artificial Analysis Intelligence Index v4.3.x, which Meyer describes as “a map of general capability, not a verdict on your workload,” advising readers to shadow-test before switching models. At the top of the field, Opus 5.5 (released 22 September) scores 58 at max effort at $5.98 per task, or about 17 tasks per $100. At the bottom, GPT-6 Luna scores 37 at $0.07 per task — roughly 1,429 tasks per $100 — making it suited to classification, extraction and routing.

Three findings stand out in the data. First, Opus 5.5 outscores its more expensive sibling Fable 5.1 by 5 points while costing less per task ($5.98 versus $7.63). Second, Sonnet 5.5 at max effort costs more per task than Opus at max while scoring 2 points lower, which Meyer says makes it hard to justify at that setting. Third, GPT-6.1 Sol costs about one-eighth of GPT-6 Astra and one-twentieth of Fable per task for a score only 1 to 2 points lower.

The review also shows that the effort setting moves cost more than model choice. On Opus 5.5, going from xhigh to max adds 2 index points and 73% more cost per task; from medium to max, cost rises 4.46x for 7 points. Per-1M-token prices cited: Opus 5.5 at $4/$20 (input/output, cache reads $0.20), Fable and Astra at $10/$50, GPT-6.1 Sol at $2/$10, and Luna at $0.10/$0.50.

Why Cheap Review Passes Change Workflows

The core claim of the review is that routine cross-model review is now affordable. At $0.32 to $0.39 per task, Meyer writes, a Sol review pass is “cheap enough to be routine,” and a different model family reviewing Opus’s output is “a better check than Opus reviewing itself.” He runs it on every meaningful change.

The analysis also matters because it reframes cost discipline: cheaper tokens do not automatically mean cheaper work. Halving model price saves only 12.5% of real cost once human review time is counted, and a single extra minute of human review erases the saving, according to Meyer’s illustrative (not measured) example. This pushes attention toward effort settings, workflow fit and review cost rather than headline model price.

How GPT-6.1 Sol Fits the Field

GPT-6.1 Sol launched on 29 September 2026 at the same $2/$10 per-1M-token price as its week-old predecessor. Artificial Analysis already lists three effort levels: medium at 48 points and $0.21 per task, high at 50 and $0.32, and xhigh at 51 and $0.39. Even the medium setting matches the earlier GPT-6 Sol’s score of 48 at one-fifth of its $1.06 per-task cost.

Sol is notably concise — the high setting used 25M output tokens on the index against a median of 82M for comparable models. The trade-offs are latency and ceiling: high and xhigh take 57 to 69 seconds to first token, and Opus 5.5 still leads Sol xhigh by 5 points (56 versus 51). Other recent releases in the stack: Fable 5.1 (1 September), GPT-6 Astra (3 September), Opus 5.5 and Luna (22 September), and Sonnet 5.5 (28 September).

“In four weeks, the AI frontier stopped being a leaderboard and became a price curve.”

— Thorsten Meyer, ThorstenMeyerAI.com

Gaps in the Current Data

Artificial Analysis has not yet published low or max effort settings for GPT-6.1 Sol, and Meyer notes that one index point is inside the noise, so small score gaps between Sol, Astra and Fable should not be over-read. The human-review cost example is explicitly labeled illustrative, not measured. Whether Sol’s settings and pricing remain stable after launch week is unknown, and the stack reflects one practitioner’s workload — generalising it to other use cases is untested.

What the Stack Watchers Are Waiting For

Expected next steps include Artificial Analysis publishing the remaining Sol effort settings (low and max), firmer data on Sol’s real-world latency at interactive settings, and any pricing responses from Anthropic or Google to Sol’s cost position. Meyer’s stated practice is to keep Opus at high or xhigh for building, Sol for detail work and review, with Astra, Fable, Sonnet 5.5 and Luna reserved for specific jobs — a split he revisits as new benchmarks and prices land.

Key Questions

Which model does the review recommend as a default for software development?

Claude Opus 5.5 at high effort — 54 index points for $1.82 per task — for features, APIs, multi-file work and refactors, with xhigh (56 points, $3.46) reserved for architecture, migrations and trust-boundary problems.

Why use GPT-6.1 Sol instead of a higher-scoring model for review?

Sol xhigh scores 51 — only 1 to 2 points under Astra and Fable — at $0.39 per task instead of $3.26 or $7.63, making a review pass on every meaningful change affordable. It is also from a different model family than Opus, which provides a more independent check.

What are GPT-6.1 Sol’s main drawbacks?

At high and xhigh settings it takes 57 to 69 seconds to produce a first token, so it is not interactive at those levels, and Opus 5.5 still leads it by 5 index points at xhigh.

Is a higher effort setting always better?

No. On Opus 5.5, max adds only 2 points over xhigh for 73% more cost per task. Sonnet 5.5 at max jumps from $2.74 to $7.60 per task for 4 points, and the author calls max “rarely worth it.”

Do these index scores guarantee performance on my tasks?

No. Meyer describes the Artificial Analysis Intelligence Index as “a map of general capability, not a verdict on your workload” and advises shadow-testing models on your own work before switching.

Source: ThorstenMeyerAI.com