Product update · Free

GPT-6.1 Sol: when Work or Codex is enough (and Chat isn’t ready yet)

What’s new vs GPT-6 Sol, when Work/Codex helps or skip (Chat not yet; Free/Go not claimed), usage in plain language, and one complete free exercise. Pro compare toolkit draft (CSV, checklist, templates) available as a companion preview.

Published 2026-09-29 · Reviewed 2026-09-30 · Editorial rating 9.0/10

The short version

OpenAI announced GPT-6.1 Sol (OpenAI launch post; ChatGPT Work/Codex availability stated there). This Free guide keeps the main read problem-first for Plus+ users deciding when a Sol upgrade in Work or Codex is enough, states Chat not yet and Free/Go not claimed for 6.1 Sol, attributes OpenAI near-Astra and pricing claims, notes UI exact label UNKNOWN, folds prior Sol+Luna sibling context, and includes a complete one-page outline exercise with three full fictional sources. ElevateU did not run GPT-6.1 Sol for the sample.

What this changes in practice

Beginners need a decision aid—not DevDay cheerleading. The useful path is: know what OpenAI claims is new vs GPT-6 Sol, try Work/Codex Sol when Plus+ access exists, know when Chat-only/Free/Go or Astra still applies, and practice one checkable outline from messy notes on any model.

What’s new

On 2026-09-29, OpenAI introduced GPT-6.1 Sol (launch post (https://openai.com/index/introducing-gpt-6-1-sol/)). A DevDay recap URL exists as secondary context but was not independently re-fetched for this preview pass—product claims below follow the primary launch page.

In everyday terms, OpenAI says you can expect:

- An upgrade path from GPT-6 Sol — positioned as stronger Sol-class help for coding, documents/PDFs, multi-step workflows, and computer use (provider framing; ElevateU did not retest). - “Near-Astra” intelligence at lower Sol list economics — OpenAI claims near-Astra intelligence at about 1/5 Astra standard API in/out token prices. That is an OpenAI claim, not ElevateU proof. - Where it shows up today — ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu starting 2026-09-29. Not yet available in Chat. - API id and list prices — gpt-6.1-sol at $2 / 1M input, $0.10 / 1M cached input, $10 / 1M output (provider). - Ultrafast (upcoming only) — OpenAI says GPT-6.1 Sol Ultrafast is coming in days (up to 8× faster token generation vs standard in Codex, per launch). Treat as not available yet—do not plan as if it is live today. - Exact UI picker label — UNKNOWN without ElevateU hands-on. Use whatever label your Work or Codex menu shows after you re-check.

Sol + Luna context: The earlier GPT-6 Sol and GPT-6 Luna (https://openai.com/index/introducing-gpt-6-sol-and-luna/) post (2026-09-22) introduced Sol as the main Work/Codex Sol-class pick and Luna as a cheaper/faster tier (Free/Go Luna in the desktop app per that post; Sol/Luna family still not in Chat). This guide folds that story into the 6.1 Sol decision path. 6.1 Sol Free/Go ChatGPT access is still not claimed in the 6.1 launch post (unlike Luna desktop for base Luna). If you only have Free/Go, do not assume 6.1 Sol appears.

Attribution rule: All “near-Astra,” speed, cost %, and eval numbers on OpenAI pages are OpenAI or partner claims. ElevateU did not run independent benchmarks or a live GPT-6.1 Sol session for this guide.

Astra context: OpenAI still positions hardest work toward Astra (Astra page (https://openai.com/index/gpt-6-astra/)). Use Astra-class help when that is what your job needs—and you actually have access. This guide is about when Work or Codex Sol is enough, not a substitute for an Astra decision.

Where this could help

Try GPT-6.1 Sol when you are on a Plus+ plan that can open ChatGPT Work or Codex, the job is everyday or well-scoped professional work you will still check yourself, and 6.1 Sol (or a Sol-class label) appears in your menu.

Three concrete situations:

Situation — Why Work/Codex Sol might help (OpenAI’s framing) Iterate on code or a well-scoped fix inside Codex / Work — Stronger Sol-class coding help without always reaching for Astra rates (provider) Turn a messy PDF, doc pack, or multi-file brief into a structured draft — Docs/PDFs and multi-step workflows OpenAI positions for Sol-class upgrades Run a multi-step work handoff (notes → outline → next actions) you can verify — Everyday professional workflows in Work/Codex at Sol list economics (provider)

You still own the check. A cleaner outline is only useful if the facts match your sources.

What it does not replace

Stay on your current model (or wait) when:

- You only have Chat today—or you are on Free / Go with no claimed 6.1 Sol access. Do not assume ChatGPT Chat or Free/Go will open GPT-6.1 Sol. Chat is not yet available for this model (OpenAI). - The work is hardest / deepest judgment or computer-use OpenAI still points to Astra—and you actually have Astra-class access for that job. - Your current model already finishes the task well enough—switching for novelty alone wastes time. - You are switching only because an OpenAI benchmark or “near-Astra” phrase looks better—that is not ElevateU proof. - Stakes are high (legal, medical, financial, customer commitments)—you still review the result, whichever model you use.

This is beginner editorial guidance, not a lab ranking of model quality.

Access and cost

- About 10–15 minutes - Any chat / Work / Codex assistant you can use at work (6.1 Sol optional) - The three sample sources below (or three real messy notes of your own)

Remember: A ChatGPT subscription is not unlimited use. API token meters are not the same as Work/Codex usage limits. Free/Go access to 6.1 Sol is not claimed in the launch post.

Understand usage and cost before you start

Remember: A ChatGPT subscription is not unlimited use. API token meters are not the same as Work/Codex usage limits. Free/Go access to 6.1 Sol is not claimed in the launch post.

- Work / Codex (provider) — Plus, Pro, Business, Enterprise, Edu: GPT-6.1 Sol available in Work and Codex starting 2026-09-29 (OpenAI launch). Chat: not yet. Exact UI label = UNKNOWN—confirm in your menu. - API tokens ≠ chat plan — If you build with the API, list prices are $2 / $0.10 cached / $10 per 1M in/cached/out for gpt-6.1-sol (provider). That meter is separate from Plus/Pro Work/Codex usage. - “Near-Astra at ~1/5 Astra standard API prices” — OpenAI claim about API in/out economics vs Astra. Attribute to OpenAI; ElevateU did not verify your bill. - Ultrafast — Coming days (provider)—not yet for budgeting as live capacity. - Free / Go — Do not promise Free or Go ChatGPT access to GPT-6.1 Sol. Prior Luna desktop Free/Go access (Sol/Luna post) is not the same as 6.1 Sol. - No invented burn rates — ElevateU does not invent “one outline = X messages” or “one Codex run = $Y.” Re-check live pricing and Usage before you budget.

For plan surfaces and API id, see the optional sections at the end.

Optional: try a related example

Start with the fictional example. No ElevateU membership is required.

You will turn three short, messy sources into a 1-page outline with four headings: Problem / Done / Next / Open questions. Practice this on any chat or Work assistant you are allowed to use. If GPT-6.1 Sol appears in Work or Codex, you may run the same prompt there for a compare—you do not need 6.1 Sol to learn the check. Free/Go and Chat-only readers can still complete the skill on whatever model they already have.

Discipline to practice: do not invent decisions, owners, dates, or budget numbers. If something is not in the sources, put it under Open questions or write Not stated.

What you need

- About 10–15 minutes
- Any chat / Work / Codex assistant you can use at work (6.1 Sol optional)
- The three sample sources below (or three real messy notes of your own)

Step A — Read the three complete sample sources

Source 1 — Voice-memo dump from Jordan Lee (ops lead), 29 Sep 2026

Quick dump — Cedarline Tutoring Co-op, fall parent orientation packet.

Board contact is Marcus Chen. Kickoff was Monday 28 Sep. I told Marcus we'd send a one-page outline this week: problem, what's already done, what we do next, open questions. I did NOT promise a print date or a budget number.

From the call:
• Biggest pain is the schedule sheet — room numbers change in a shared Doc and the PDF parents get still has last year's rooms half the time.
• Registration form on mobile cuts off the emergency-contact field; two parents emailed they couldn't finish on their phones.
• Keep the cedar-green brand; no full rebrand.

I have not assigned who owns the schedule-sheet fix. Do not invent an owner.

Source 2 — Slack DM, #cedarline-ops, 30 Sep 2026

Priya Nair  8:41 AM
Jordan — inventory update for orientation.

Done already:
• Folder of 12 orientation handouts (Drive: Cedarline-orientation-2026-09-28).
• Phone screenshots of Registration on Pixel 7a (same folder).
• Confirmed cedar-green hex on letterhead matches CSS --cl-cedar.

Not done:
• Schedule-sheet room sync process for tutors.
• Registration mobile field fix mock.

Please don't tell Marcus we have a print date for the parent PDF. We don't. Also nobody owns tutor schedule training yet.

Source 3 — Email fragment from Marcus Chen (Cedarline board), 28 Sep 2026

From: Marcus Chen <[email protected]>
To: Jordan Lee <[email protected]>
Subject: Re: Monday kickoff — what the board needs in writing

Jordan,

Thanks for the call. For the board packet I only need a short outline:
1) the problem in plain language
2) what you've already finished
3) what you'll do next
4) open questions for us

Please do not invent a print budget or a print-week. The board has not approved either.

— Marcus

Step B — Copy this full prompt

You are helping me write a one-page project outline for a nonprofit board packet.

Use ONLY the three sources I paste below. Do not invent owners, print dates, budgets, print-weeks, or decisions that are not stated.

If something is missing, put it under Open questions or write: Not stated.

Output exactly these four sections, in this order, with short bullets:
1) Problem — what hurts, for whom, in plain language
2) Done — what is already finished (dates/files only if stated)
3) Next — concrete next work; owners only if named; otherwise Not stated
4) Open questions — anything the board still must answer

Rules:
- Keep the whole answer under 220 words.
- Calm, factual, no sales fluff.
- Audience: board packet via Marcus — no fake promises.
- End with a one-line Sources used list.

SOURCES:
[paste Source 1]
[paste Source 2]
[paste Source 3]

See the complete example and what to check

Step C — Illustrative full response

The block below is an example of a good answer shape. ElevateU did not generate this from a live GPT-6.1 Sol session; it is written to match the instructions above so you can see what “done” looks like. You can practice the same check on any model.

1) Problem
• Cedarline Tutoring Co-op’s parent orientation materials are unreliable: schedule-sheet room numbers change in a shared Doc while the parent PDF still shows last year’s rooms (Jordan voice memo).
• Registration on mobile cuts off the emergency-contact field; two parents emailed they could not finish on phones (Jordan voice memo).
• Goal of this outline: problem / done / next / open questions for Marcus’s board packet — no print date promised (Jordan; Marcus email).

2) Done
• Folder of 12 orientation handouts (Drive: Cedarline-orientation-2026-09-28) (Priya Slack).
• Phone screenshots of Registration on Pixel 7a (same folder) (Priya Slack).
• Confirmed letterhead cedar-green matches CSS variable --cl-cedar (Priya Slack).
• Kickoff Monday 28 Sep; cedar-green brand kept; no full rebrand (Jordan voice memo).

3) Next
• Send this one-page outline to Marcus this week (Jordan voice memo).
• Schedule-sheet room sync process for tutors — owner Not stated (Priya Slack).
• Registration mobile field fix mock — owner Not stated (Priya Slack).

4) Open questions
• Who owns the schedule-sheet fix / tutor schedule training? (not assigned — Jordan; Priya)
• Board-approved print budget? Not stated — Marcus asked us not to invent one.
• Board-approved print-week? Not stated — Marcus asked us not to invent one.

Sources used: Jordan Lee voice memo 29 Sep 2026; Priya Nair Slack 30 Sep 2026; Marcus Chen email 28 Sep 2026.

Step D — How to check your own result

Use this checklist on your model output (everyday model and GPT-6.1 Sol, if visible in Work/Codex):

1. Source lock — Every fact appears in at least one pasted source. If you cannot point to the line, delete it or mark Not stated / Open questions.
2. No invented owners — Schedule-sheet sync and Registration mock must stay Not stated (or open questions)—do not assign Jordan, Priya, or Marcus as owner of unfinished work unless a source names them.
3. No invented print budget or print-week — Marcus explicitly forbade inventing print budget or print-week.
4. Four headings present — Problem / Done / Next / Open questions in that order.
5. Done vs next — Handout folder, screenshots, and cedar-green hex belong in Done; schedule sync and Registration mock belong in Next.
6. Audience — Board-packet outline for Marcus—not a sales pitch or a fake timeline email.
7. Compare A vs B (optional) — If you run the same prompt on two models (e.g. prior Sol vs 6.1 Sol in Work/Codex, or your usual Chat model), prefer fewer invented claims and less cleanup—not the longer answer. If 6.1 Sol is not visible, you still completed the skill.

If GPT-6.1 Sol never appears (Chat-only, Free/Go, or menu still updating), you still practiced a strict, checkable one-page outline from messy sources. Re-check Work/Codex; ElevateU cannot name the exact UI string without hands-on and does not promise Free/Go or Chat access to 6.1 Sol.

Illustrative teaching example; not a recorded model response. Review the result before using real information.

Companion toolkit / Pro

GPT-6.1 Sol compare toolkit

Compare CSV, evaluation checklist, Templates A/B/C + variants, Cedarline twin, Free≠6.1 Sol / Chat-only stop paths, Work/Codex UI label UNKNOWN, Astra compare only when visible — Free basics stay free. No ElevateU Sol scores.

Expected result + requirements (everyday labels)
First exercise + templates A/B/C + short variants
CSV: EXAMPLE teaching then WORK blanks; model_label real on WORK
Usage per-run only (balance ≠ burn); Free≠6.1 Sol / Chat-only stop; Work/Codex UI UNKNOWN
Evaluation checklist mirroring Free Step D (Cedarline)
Illustrative bad vs good Cedarline outline case (not Sol live run)

The free exercise above is complete. Pro adds reusable templates, a worked case, an editable worksheet and troubleshooting.

Explore the toolkit →

Members with Pro access can open the full materials. New paid subscriptions are not open yet. ElevateU Pro does not include an AI provider subscription.

Sources and review notes

Introducing GPT-6.1 Sol — OpenAI ↗

OpenAI — Introducing GPT-6.1 Sol (PRIMARY) ↗

OpenAI — DevDay 2026 recap (secondary; not re-verified in this preview pass) ↗

OpenAI — GPT-6 Sol and GPT-6 Luna (sibling context) ↗

OpenAI — GPT-6 Astra (Astra context only) ↗

ElevateU editorial brief. Capability and near-Astra claims are OpenAI’s unless labeled otherwise. Chat not yet; Free/Go not claimed for 6.1 Sol. Exact Work/Codex UI picker label = UNKNOWN without hands-on. ElevateU did not run a live GPT-6.1 Sol session for the sample exercise.

Product claims and speaker forecasts belong to their sources. Our examples are editorial interpretations. Last reviewed: 2026-09-30.