AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: My September 2026 AI Workflow: Build With Opus, Dig With Sol on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

In a September 29 account, Thorsten Meyer described a workflow that uses Claude Opus 5.5 for building and newly released GPT-6.1 Sol for detailed review. He cites Artificial Analysis benchmark figures showing much lower per-task costs for Sol, while warning that benchmark scores do not establish which model will work best for a particular task.

Thorsten Meyer said September 29 that he uses Claude Opus 5.5 for building software and the newly released GPT-6.1 Sol for detailed investigation and review, a division based on benchmark capability scores and reported per-task costs. The account describes one user’s workflow; its figures do not establish which model performs best across other teams’ workloads.

Meyer bases his comparisons primarily on the Artificial Analysis Intelligence Index v4.3.x, which he describes as a general capability measure rather than a verdict on any particular workload. In his table, Opus 5.5 scores 58 at its top setting and costs $5.98 per task. GPT-6.1 Sol scores 51 at xhigh and costs $0.39 per task. The comparison spans different effort settings and does not by itself show how the models perform on the same real-world job.

For his own development work, Meyer says he uses Opus at high effort for routine building tasks and xhigh for harder work such as architecture, migrations and trust boundaries. His cited index figures put Opus at 54 for $1.82 per task at high, and 56 for $3.46 at xhigh. He says max effort costs $5.98 per task for a score of 58, a trade-off he rarely considers worthwhile.

Meyer assigns Sol a different job: examining a specific file or code change and providing a second review. Artificial Analysis figures cited in the account list Sol at 50 points and $0.32 per task at high, and 51 points and $0.39 at xhigh. The account also reports first-token times of 57 and 69 seconds at those settings, respectively, which may make them unsuitable for interactive use. Meyer says Astra, Fable, Sonnet and Luna remain options for more specific tasks, while Jev handles high-volume yes-or-no and routing decisions.

At a glance
reportWhen: Published September 29, 2026; GPT-6.1 S…
The developmentThorsten Meyer published a September 29 account of using Opus 5.5 for development and GPT-6.1 Sol for review after Sol’s release that day.

Opus builds. Sol reviews. Jev decides.

The September 2026 AI stack in one page: six frontier models on one price curve, and a decision model for the high-volume judgements that do not need a sentence.
Scores: Artificial Analysis Intelligence Index v4.3.x. Data as of 29 September 2026.
BuildsClaude Opus 5.5 at high or xhigh effort
Digs and reviewsGPT-6.1 Sol at high or xhigh effort
DecidesJev on high-volume yes/no and routing calls

One price tape, six models

Put every model on the same cost-per-task ruler and capability looks compressed. The bill does not.
Price tape: cost per task of six models on a log scale, from GPT-6 Luna at $0.07 to Fable 5.1 at $7.63$0.05$0.10$0.50$1$5$10cost per task, log scale: each tick is a different order of magnitudeGPT-6 Lunaindex 37 · $0.07GPT-6.1 Solindex 51 · $0.39 (xhigh)GPT-6 Astraindex 53 · $3.26Opus 5.5index 58 · $5.98Sonnet 5.5 · index 56 · $7.60Fable 5.1 · index 53 · $7.63about 100× from the cheapest to the priciest, but only 21 index points between them

Score against cost, at every effort setting

Each dot is an effort level. Opus 5.5 at high already matches Astra and Fable at max on this index, for less money.
Intelligence Index score against cost per task for each effort setting of six models$0.01$0.10$1$102030405060cost per Intelligence Index task, log scaleindexOpus high / xhigh: my defaultOpus 5.5Sonnet 5.5Fable 5.1GPT-6 AstraGPT-6.1 Sol (new)GPT-6 Sol (Sep 22), dashedGPT-6 Lunaup and to the left is better
Astra and Fable are shown at their top published setting. Luna starts at $0.0045 per task. GPT-6.1 Sol has no low or max setting published yet.

The effort dial moves the bill more than the model

Going from medium to max on Opus costs 4.46× more for 7 points. That is why I run high or xhigh.

Claude Opus 5.5

$0.55
42
$1.34
51
$1.82
54
$3.46
56
$5.98
58
low
medium
high
xhigh
max
Solid bars are where I run it. Max adds 2 points over xhigh for 73% more cost.

Claude Sonnet 5.5

$0.41
36
$0.59
41
$1.08
47
$2.74
52
$7.60
56
low
medium
high
xhigh
max
Best value is high. At max it writes about 193k output tokens per task, the most measured.

GPT-6.1 Sol: near-Astra scores at a fraction of the price

Launched 29 September at $2 in and $10 out per 1M tokens. It sits 1 to 2 points under Astra and Fable, and Opus xhigh still leads it by 5.

Three published settings

SettingIndexCost per taskOutput tokensFirst token
medium48$0.2115M5.3 s
high50$0.3225M57 s
xhigh51$0.3936M69 s
Median for comparable models is 82M output tokens. High and xhigh are not interactive: plan for a wait before the first token.

Same score band, very different bill

GPT-6.1 Sol xhigh
$0.39index 51
Opus 5.5 high
$1.82index 54
GPT-6 Astra max
$3.26index 53
Opus 5.5 xhigh
$3.46index 56
Fable 5.1 max
$7.63index 53
Cost per Intelligence Index task. A one-point gap is inside the noise.

My stack: who builds, who reviews

Opus does the work. A second model family reviews it, because a different reviewer catches what the author cannot see.
Stack diagram: Opus 5.5 builds at high effort, escalates to xhigh, and sends every change to GPT-6.1 Sol for review; Astra or Fable give a second opinionOpus 5.5 · xhighhard problems: architecture,migrations, trust boundariesOpus 5.5 · highMAIN BUILDERfeatures, APIs, multi-filework, refactorsescalate when it gets hardGPT-6.1 Solhigh or xhighdigs into details andreviews every change$0.32–0.39 per taskdifffindingsAstra or Fablesecond opinion, 8 to 20×the cost per taskif they disagreeSonnet 5.5 · Lunaside work: scopedsubtasks, bulk checksand routingFailed review? Hand Opus the failing case and the evidence.Never just “try harder”: effort cannot supply a missing requirement.
Effort is not capability. Turning the dial up does not make a model smarter.
Effort cannot fill gaps. A missing requirement stays missing at any setting.
Different model, same spec. That is not independent review if both read the same flawed brief.
Green tests are not approval. Passing tests only prove what the tests cover.

Cheaper tokens are not cheaper work

Illustrative, not measured: $1 of model time plus 4 minutes of review at $45 an hour. Halving the model price saves 12.5% of the total. One extra minute of review erases it.
$4.00
review $3.00
model $1.00
Baseline
$3.50
review $3.00
model $0.50
Model price cut 50%
$4.25
review $3.75
model $0.50
Cheaper model plus 1 extra minute of review
Track cost per accepted result: model, tools, review and rework, divided by the results someone actually uses.

Read the numbers with four warnings

The index movesFable scored 66 on an earlier version and 53 on v4.3. Compare within one version only.
Fallback is includedFlagged cyber and biology tasks route to older Anthropic models, now on Sonnet 5.5 too.
Max is not productionReal deployments run medium or high, where gaps narrow and costs fall.
Your work decidesShadow-test on your own tasks. Budget cost per task, not per token.

Part 2: Jev, the model that decides instead of writing

Jev cannot write, summarise or extract. It answers narrow typed questions with a probability and an honest confidence, in under a second, for about $0.04 per million input tokens.

One call in, typed answers out

Your code, not Jev, decides what to do with each answer, usually by confidence band.
Jev flow: state and typed questions go into one Jev call; typed answers with confidence come out; code acts alone, escalates the gray zone, or logsStatea ticket, a story,a site profile,a log line …+ typed questions,many per callJevone call0.3 to 0.9 s$0.042 / M tokens inAnswersnoul: 0.03choice: billing p 0.91, conf 0.86score: 2.7 of 3 conf 0.64code branches on thisAct aloneconf ≥ 0.8Escalategray zone toLLM or humanLogmeasure first

Three question types

noul
A yes/no question. Returns the probability of yes, 0 to 1.
gates, flags, filters
choice
Pick one option. Returns the choice, a probability per option, and a confidence.
routing, classification, taxonomy
score
Rate on your ordered levels. Returns a position (it can fall between levels) plus a confidence.
quality, fit, severity, priority

Confidence is the superpower

In my own measurement on a 31-topic classification, Jev agreed with a frontier LLM almost every time it was sure, and rarely when it was not. So: decide the clear cases, route the gray zone.
confidence 0.8 or higher
97–99%
all answers
89%
confidence below 0.5
42%
Agreement with a frontier LLM, my production data, September 2026, rounded.

Three uses running in my publishing operation

About 90,000 decisions so far. Checks I could only afford on a sample now cover everything.
$2.01
Language check
78,889 articles scanned overnight. 1,576 in the wrong language found, 1,553 fixed in place.
22%
Relevance gate
About 10,000 story-to-site pairings judged in 3 days. Only 22% were clearly on-topic.
89%
Classifier fallback
Agreement with the primary LLM across 31 topics, used when that LLM errors.

The fit test, then the shadow test

Use Jev only when all four hold. Then prove it on past decisions before it acts on anything.
High volumeThousands of small calls, not a handful of big ones.
Narrow questionNo multi-step reasoning needed.
Cheap errorsOr unsure cases go to something smarter.
Heuristic failsVisibly, and measured, not assumed.
  1. Replay 300 to 500 past decisions
  2. Compare overall and per confidence band
  3. Read 20 disagreements, decide who was right
  4. High band at 95% or better?
  5. Own flag, off by default
  6. Canary on 5 to 10 units
  7. Roll out in the confident band only

24 use cases, sorted by how well they fit

Start from the strong fits. The amber ones need a measurement before you trust them, and the red ones fail one of the four conditions.
in productionstrong fitmeasure firstpoor fit

Proven in production

  • 1Relevance gate
  • 2Language check
  • 3Classifier fallback

Publishing and content

  • 4Thin-source detector
  • 5Same-event dedupe
  • 6Product fits roundup
  • 7Disclosure present
  • 8Headline quality
  • 9Comment moderation

Commerce and support

  • 10Support-ticket routing
  • 11Return-reason coding
  • 12Review to feature complaints
  • 13Catalogue taxonomy
  • 14Order-fraud pre-triage

Software and AI systems

  • 15LLM guardrail
  • 16RAG passage filter
  • 17Citation check
  • 18Tool and intent routing
  • 19Log-line triage
  • 20PR risk triage

Business ops and home

  • 21Inbox triage
  • 22Expense categorisation
  • 23Lead qualification
  • 24Smart-home intent

Limits, cost and one hard rule

No writing, summarising or extractionPair it with an LLM for the write step.
No world knowledgePut a snippet in the state; a bare name means nothing.
Reads your wording literallyA rewording moved my results about 2 points. Freeze it, re-measure after changes.
Weaker on non-English, maths, datesKeep those checks on an LLM. Early access, hosted API only.
100,000 decisions ≈ $2.50
About 60M input tokens at $0.042 per million, output free, roughly 600 tokens per three-question call. Latency 0.3 to 0.9 seconds.
Never the sole decision-maker for consequences about people. Hiring, credit, medical and legal outcomes stay with a human. Jev can sort and flag. A person decides.
Sources. Model scores, cost per task and speeds: Artificial Analysis, Intelligence Index v4.3.x, including the GPT-6.1 Sol medium, high and xhigh pages, checked 29 September 2026. Astra and Fable scores from the Artificial Analysis v4.3 announcement. Jev figures are my own production measurements, September 2026, rounded. The review-bill example is illustrative. Read the full article on thorstenmeyerai.com.

Why the Review Seat Matters

Meyer’s proposed split makes model selection a question of cost per useful task, not just a ranking by benchmark score. If a lower-cost model can provide a helpful review, running it across more changes may be affordable. The account’s figures suggest a sizeable price difference between Sol and several higher-scoring options, but they do not measure the quality or time savings of Meyer’s actual review process.

The workflow also highlights a limit of automated checks: Meyer argues that a second model can catch problems, but that model may still share a flawed specification or miss a failing case. He says passing tests are not approval to ship, and recommends giving the builder concrete failure evidence when review finds a problem. These are his process recommendations, not independently evaluated results.

Products Worth Considering

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Kindle Paperwhite Signature Edition 32GB (2024 Model) - Metallic Black

Kindle Paperwhite Signature Edition 32GB (2024 Model) – Metallic Black

As an affiliate, we earn on qualifying purchases.

From Model Rankings to Task Costs

Meyer frames his September comparison as a shift from leaderboard positions toward the cost of completing a task. He says six models fall within about 20 index points of one another, while their reported task costs differ by roughly 100 times. The source does not specify the full calculation behind that broad range, so it should be read as his summary of the table rather than a universal price ratio.

His account says Opus 5.5 was released September 22, Sonnet 5.5 on September 28, and GPT-6.1 Sol on September 29. It reports that Sol kept the listed token prices of its predecessor, GPT-6 Sol, at $2 per million input tokens and $10 per million output tokens. These release dates, prices and benchmark readings are those reported in the source and may change over time.

The figures also depend on the effort setting. Meyer’s table shows Opus rising from 51 index points and $1.34 per task at medium to 58 and $5.98 at max. He reports that Sonnet at max generated about 193,000 output tokens per task in the benchmark, while its listed task cost rose from $2.74 at xhigh to $7.60 at max. Such benchmark task costs are not the same as a team’s total labor or operating costs.

“The practical reading: Sol is not the model I ask to build. It is the model I can afford to run on everything.”

— Thorsten Meyer

What the Benchmarks Cannot Show

The source does not provide independent testing of Meyer’s workflow, the underlying task prompts, or evidence that the benchmark costs match typical software projects. It also cautions that one index point may fall within measurement noise and says Artificial Analysis had not published Sol’s low or max settings at the time of writing. The account does not establish whether the models’ relative performance holds for other tasks, users or later versions.

Some comparisons use different settings, and the reported cost per task reflects benchmark usage rather than the full cost of engineering work. Meyer offers an illustrative point that human review time can outweigh token-price savings, but the source excerpt ends before providing the full example. The amount of human time saved, the review’s error rate and the total cost of the proposed stack remain unclear.

Testing the Stack on Real Work

Meyer advises teams to shadow-test models on their own work before changing defaults. That means comparing outputs against the existing process and a defined quality bar before routing more tasks to a new model. The source does not announce a formal trial, publication schedule or additional benchmark release.

For readers considering the same division of labor, the next useful evidence would be results from their own tasks: whether Sol catches meaningful issues, how much review time it takes, and whether Opus at high or xhigh meets the required standard. Updated index scores, newly published Sol effort settings and measured end-to-end costs could change the comparison; none is confirmed in Meyer’s account as a future date or commitment.

Key Questions

What workflow did Thorsten Meyer describe?

He said he uses Opus 5.5 for building and GPT-6.1 Sol for detailed investigation and review, with other models assigned to specific tasks.

Why does Meyer use GPT-6.1 Sol for review?

He points to its reported benchmark cost of $0.32 per task at high and $0.39 at xhigh, and says the lower cost makes routine review practical for his workflow. The source does not independently measure whether that review improves outcomes.

Does the comparison prove Sol is better value for every team?

No. The index measures general capability, and the source says teams should shadow-test on their own workloads. Benchmark costs and scores may not reflect a team’s actual tasks or review time.

What trade-off does the source report for Sol at higher effort?

At high and xhigh, the cited first-token times are 57 and 69 seconds. Meyer says that delay makes those settings unsuitable for interactive work.

Source: ThorstenMeyerAI.com

You May Also Like

Why Premium Kitchen Appliances Win on Convenience More Than Hype

The truth about premium kitchen appliances’ convenience lies beyond the hype, revealing how smart features genuinely transform your cooking experience.

Why Software Support Matters as Much as Hardware

Learn why software support is just as vital as hardware for maintaining security and performance, and discover how to keep your systems running smoothly.

Amazon’s AI Shopping Overhaul

Amazon has introduced a significant overhaul of its shopping platform, integrating advanced AI features to enhance user experience and personalization.

USB-C Explained Without the Confusion

Here’s a clear, straightforward guide to understanding USB-C without the confusion, so you can simplify your tech setup and stay ahead.