OpenAI shipped GPT-6.1 Sol on 29 Sep 2026 at DevDay — about a week after GPT-6 Sol. Pitch: near-Astra on agentic coding, computer use, and professional work at one-fifth Astra’s standard API input/output prices. Same $2 / $10 headline as Sol; cached input drops to $0.10 (half Sol’s $0.20). If you already bill Work, Codex, or API against gpt-6-sol — or parked hard briefs on Astra — the question is flip the model ID, skip Astra for this job, keep Astra when the 5× rate still earns the seat, or wait.
This is a switch / skip Astra / keep Astra / wait sheet. Not a rematch of GPT-6 Sol vs Luna for freelancers. Not a rewrite of Astra Fast, Pro caps, or ChatGPT allowances — those stay on the Astra when-to-pay post. Cross-vendor $/MTok stays on the API cost ladder. Here: does gpt-6.1-sol replace Sol as your workhorse — now, later, or not at all?
Affiliate disclosure: some tool or platform links may later be affiliate or referral. Prices below are from OpenAI’s introducing post and live API model docs as of 30 Sep 2026 (SAST) — re-check before you quote a client. Not legal advice; not a guarantee of clients, savings, or income.
What shipped 29 Sep 2026 — GPT-6.1 Sol in plain English
On 29 Sep 2026, OpenAI published Introducing GPT-6.1 Sol. It is an upgrade to GPT-6 Sol. The API model ID is gpt-6.1-sol on the GPT-6.1 Sol model page. OpenAI’s own framing: it nearly matches GPT-6 Astra on agentic coding, computer use, and professional work at one-fifth of Astra’s standard input and output token prices. Treat “near-Astra” and “one-fifth” as vendor claims on their evals and list rates — not a Mr1Tech guarantee that every client job will bill five times less or clear Astra-grade acceptance.
Availability on day one: Plus, Pro, Business, Enterprise, and Edu in ChatGPT Work and Codex, plus the API. It is not yet in Chat. Context window is 1,050,000; max output 128,000; knowledge cutoff 30 Apr 2026. reasoning.effort supports low, medium (default), high, xhigh, and max — not none or minimal. Free API tier: not supported. Tool-specific fees (search, computer use, and friends) may add per-call charges on OpenAI’s pricing page — do not invent tool dollars here.
OpenAI did not launch GPT-6.1 Astra at DevDay. One calm fact for the queue; this sheet is about the Sol upgrade, not a fear cycle. For Luna-shaped cheap clerical volume, keep using the Sol vs Luna chooser — do not rematch tiers here.
Same $2/$10 headline, cheaper cache
Two numbers get mashed together in secondary posts. Keep them separate.
List price (Standard, per 1M tokens). GPT-6.1 Sol: $2 input / $0.10 cached input / $2.50 cache writes / $10 output — cite the introducing post and gpt-6.1-sol. GPT-6 Sol on gpt-6-sol is still $2 / $0.20 / $2.50 / $10. Astra on gpt-6-astra is still $10 / $1 / $12.50 / $50. The uncached $2 / $10 pair did not move versus Sol. The cache row did: $0.10 vs Sol’s $0.20 — half the story for long system prompts, policy packs, and repo prefixes that reuse across agent turns. Cached input on 6.1 Sol is priced at 5% of uncached input; cache writes at 1.25× uncached input.
“One-fifth” Astra. That is OpenAI’s framing on standard I/O versus Astra’s $10 / $50 pair. It is not a guarantee that your invoice drops 5×. Prompts over 272K input still hit 2× input/cache and 1.5× output for the full request. Fast mode is 2× Standard. Batch and Flex are 50% of Standard. Regional processing adds +10% where available. Quote the USD rows, then meter one real job. Do not invent personal bill savings, ZAR conversions, or ChatGPT message quotas.
Switch from GPT-6 Sol now — when the upgrade is boring and correct
The upgrade is boring and correct when your week is already Sol-shaped on API, Codex, or ChatGPT Work: everyday coding, debugging, docs, and multi-step workflows you already ship against gpt-6-sol. Agent loops that reuse a long cached prefix — system prompt, policy pack, repo context — feel the $0.10 vs $0.20 cached-input cut every turn. Same uncached list price means you are not waiting for a discount; you are checking whether OpenAI’s quality bump lands on your acceptance criteria.
Meter one non-secret job at default medium reasoning effort before flipping production defaults. Same brief, same acceptance you already cleared on Sol. Log token bill, rework minutes, wall time, and whether cache hits showed up on the meter. If the deliverable still passes review and the meter is quieter or flat, swapping the model ID is the low-drama move.
OpenAI’s introducing post cites vendor benches on SWE-style coding, PDF professional Q&A, multi-tool business workflows, computer use, and science. Useful direction. Quote them as OpenAI claims, not win-rates you paste into a client proposal. Skip inventing “average freelancer Sol bills.” One logged row beats a launch chart. Joburg freelancers often meter API and Codex runs in USD on a card while quoting clients in rands — that packaging alone is why a same-job A/B before flipping defaults matters. Do not invent FX markups or an “SA average Sol bill.”
Skip Astra (for this job) — when 6.1 Sol is enough
OpenAI itself tells you to compare Sol versus Astra on your tasks for the quality-versus-cost tradeoff. Freelancer rule: well-scoped coding, computer-use, and professional workflows where last week’s Astra sheet would have said “maybe try the premium row” now get a cheaper first try on gpt-6.1-sol. At Standard list, 6.1 Sol’s $2 / $10 sits at one-fifth Astra’s $10 / $50 on OpenAI’s framing — so the wallet default for a first meter is the Sol upgrade, not a vanity Astra flip.
Keep Astra when the brief is hardest scientific or open-ended judgment work, when your one-job A/B on 6.1 Sol still fails acceptance, or when the client already pays for fewer premium loops and you can defend a hard USD spend cap. That when-to-pay logic lives on the complementary Astra when-to-pay sheet. Do not rewrite Astra Fast mode, Pro caps, or tool-fee math here. Agents harness meters stay on the Agents API freelancer budget post — model ID swap is only one layer.
Skip Astra for this job when the SOW is well-scoped implementation after design is locked, a long cached agent prefix already dominates the bill, a PDF/doc professional pack clears review on Sol-tier quality, or a multi-step business workflow with tools already ships inside budget on a metered 6.1 Sol run. Keep Astra when rework minutes on Sol would cost more than the 5× list gap — still framed as a logged row, not a launch slogan.
Wait / stay — when not to touch the default yet
Wait when you need the model in Chat (not Work or Codex) and it is not there yet — OpenAI’s introducing post is explicit. Wait when the client SOW locks gpt-6-sol or gpt-6-astra; change order first, not a silent string swap. Wait when you have not metered one job yet; a DevDay post alone is not a production flip mid-client week.
Wait if you were specifically holding for Sol Ultrafast. OpenAI says GPT-6.1 Sol Ultrafast is coming in the coming days, with up to 8× faster token generation in Codex on their framing — it is not documented as a live model endpoint at draft time (no Ultrafast model page to quote). Pro 500 / Astra Ultrafast sit on a different product ladder — one sentence max; this is not a $500 plan review.
Stay on GPT-6 Sol only if quality already ships inside budget and you refuse a one-job A/B this week. Same uncached list price means “wait for a discount” is not the reason — Chat gate, Ultrafast timeline, SOW lock, or migration caution is. If Luna already wins your clerical volume meter, keep that seat on the Sol vs Luna sheet; do not drag volume work onto 6.1 Sol just because DevDay shipped an upgrade.
Buy/skip matrix (sheet)
| Job | Switch Sol → 6.1 Sol now | Skip Astra — use 6.1 Sol | Keep Astra for hardest jobs | Wait (Chat / Ultrafast / SOW) | Stay Sol |
|---|---|---|---|---|---|
| Everyday Codex / Work coding | Yes — meter one job | Yes if well-scoped | Only if 6.1 Sol fails | If SOW locks old ID | If you refuse A/B |
| Long cached agent prefix | Yes — $0.10 cache story | Usually yes | Rarely for cache alone | — | If already shipping |
| PDF / doc professional pack | Yes — meter one job | Yes as first try | If acceptance still fails | — | If already inside budget |
| Multi-step business workflow + tools | Yes at default medium | Yes if scoped | If judgment / stakes demand it | If tools meter unclear | — |
| Computer-use session | Yes — meter one job | First try on 6.1 Sol | If Sol-tier fails stakes | — | — |
| Science / research hard case | Maybe — still meter | Weak as default | Yes when Sol-tier fails | — | — |
| Chat-only freelancer (no Work/Codex yet) | No — not in Chat yet | — | — | Yes — wait for Chat or move to Work/API | Stay on current Chat seat |
Get the 1-page Sol → 6.1 Sol / skip Astra / keep Astra / wait chooser
Optional printable of this free guide. Soft link until checkout. Not a “paid summary.” Not a guarantee of clients, savings, or income.
This week’s action — one non-secret Sol-shaped job
Pick one non-secret Sol-shaped job you already ran on GPT-6 Sol. Same brief, same acceptance criteria. Re-run it on gpt-6.1-sol in the API or in Codex / ChatGPT Work at default medium effort. Compare token bill, rework minutes, and wall time — note cache hits if any. Fill one sheet row: switch Sol → 6.1 Sol now / skip Astra use 6.1 Sol / keep Astra for hardest jobs / wait (Chat, Ultrafast, or SOW) / stay Sol.
If the job still fails acceptance, escalate to Astra per the when-to-pay row — not a panic-buy of Pro 500. Do not change a production default mid-client week on a launch post alone.
Sources: Introducing GPT-6.1 Sol, gpt-6.1-sol, gpt-6-sol, gpt-6-astra — re-check Chat availability, Ultrafast readiness, and pricing rows at paste time.