Skip to main content
AI Choice Engine
Back to blog

How-to

Migrating from GPT-5.6 to GPT-6.1 Sol and Luna — Prices, Model IDs, Caching, and What Breaks

GPT-6.1 Sol halves GPT-5.6 Sol's promotional API rate. Luna halves input and lowers output by about 58%. Compare exact IDs, caching and confirmed price guarantees before migrating.

Published September 26, 2026Updated October 1, 20267 min readAI ToolsBy AI Choice Engine Editorial

Prices and promotional terms were rechecked on 1 October 2026. On 29 September OpenAI released GPT-6.1 Sol at the same $2/$10 base rates as GPT-6 Sol, with lower cached-input pricing ($0.10 vs $0.20 per million). The API, model behavior, and ordinary ChatGPT availability are separate; verify the exact ID and surface before changing traffic.

On 22 September 2026 OpenAI released GPT-6 Sol and GPT-6 Luna; it released GPT-6.1 Sol on 29 September. GPT-6.1 is now the newer Sol option at the same $2/$10 base rates as GPT-6 Sol and $0.10 cached input. Both GPT-6 Sol versions and GPT-6 Luna are separate model IDs; GPT-6 Sol still costs $0.20 per million cached input tokens. GPT-5.6 Sol's current promotional price is guaranteed at least through 21 November 2026; no subsequent rate or definite end date is confirmed. Plan a migration and test the exact target ID rather than assuming the earlier GPT-6 Sol and the newer GPT-6.1 Sol behave identically.

This is the checklist. It covers what changes, what to re-test, and how to cut over without breaking production.

What changes: IDs and prices

Old IDNew IDOld price (in / out per 1M)New priceChange
gpt-5.6-solgpt-6.1-sol$4 / $20 (promo guaranteed at least through 21 Nov; launch $5 / $30)$2 / $10−50% vs promo, −60% input and −67% output vs launch; cached input $0.10
gpt-5.6-lunagpt-6-luna$0.20 / $1.20$0.10 / $0.50−50% input, −58% output
gpt-5.6-terra— (no GPT-6 Terra ID/rate in OpenAI's current model listings checked 1 Oct)$2 / $12—Keep the legacy route or evaluate GPT-6.1 Sol against the workload; do not assume a direct successor

What the savings look like at a steady monthly volume of 100M input and 20M output tokens:

WorkloadGPT-5.6GPT-6.1 Sol / GPT-6 LunaHypothetical return to GPT-5.6 Sol launch rates
Sol$800$400$1,100 at $5/$30; this future rate is not announced
Luna$44$20—

Those figures use short-context Standard rates and ignore caching, tools and processing surcharges. The launch-rate column is a planning scenario, not an announced November increase. GPT-6.1 Sol's cached-input rate is $0.10 per million; original GPT-6 Sol is $0.20 and GPT-6 Luna is $0.01.

What changes: caching

OpenAI shipped caching changes with GPT-6 that matter for agents and long conversations:

  • Higher cache hit rates by default, with cached reads at 90% below Standard input for GPT-6 Sol/Luna. GPT-6.1 Sol lists $0.10 cached versus $2 uncached input, a 95% reduction.
  • Changing reasoning effort or tool availability no longer breaks the cache. You can raise effort for a hard step, or switch tools on and off mid-conversation, and earlier context stays reusable.
  • Explicit breakpoints let you choose where a cached prefix ends.
  • A Prompt Caching Dashboard and a diagnostics tool show what is cached and explain missed opportunities.

Migration action: if your application reorders the system prompt, injects timestamps at the top, or rebuilds tool lists per request, fix that first. Stable prefixes support cache reuse; apply the exact model's cached-read rate. Check the dashboard in week one and investigate prompt layout, cache eligibility and workload patterns if the hit rate is low.

What can break

Price is the easy part. These are the things to re-test:

  1. Output shape. OpenAI's September release notes describe changes for GPT-6 Sol and Luna; those statements do not establish GPT-6.1 parity. Check the exact target model's outputs against parsers, length-based heuristics, and your evaluations.
  2. Your quality bar. OpenAI's launch claims about factuality and higher-effort Luna refer to its GPT-6 Sol/Luna evaluation. They do not establish GPT-6.1 results. Treat vendor claims as hypotheses and measure acceptance on your own workload.
  3. Effort settings. If you tuned a reasoning-effort level on GPT-5.6, re-tune it. The same label on a new model does not guarantee the same latency or token use.
  4. Long-context pricing and limits. GPT-6.1 Sol documents a 1.05M-token context and 128K max output; the >272K rate applies to the whole request at $4/$15, with cached input $0.20 and cache writes $5 per million. GPT-6 Sol has its own documented $4/$15 long-prompt tier and GPT-6 Luna $0.20/$0.75. Re-run your cost model for requests that cross 272K.
  5. Pinned aliases and dashboards. Search config, infrastructure-as-code, and billing dashboards for gpt-5.6-. Mixed IDs make a cost spike hard to read.
  6. ChatGPT is not the API. As checked 1 October, OpenAI's consumer pricing page lists GPT-6 models for Plus/Pro, while its Work/Codex model guide says GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna are unavailable in Chat. These first-party pages conflict; confirm the exact model in the account picker. Work/Codex access also depends on plan, workspace settings and rollout. Do not use a Chat window to judge API behavior.

A two-week migration plan

Days 1–2: inventory. List every call site, its current model ID, monthly tokens, and owner. Mark which ones have automated validators and which rely on humans spotting problems.

Days 3–5: shadow traffic. Send a copy of real requests to gpt-6.1-sol and, for volume routes, gpt-6-luna behind the same adapter, without serving the result. Compare validator pass rates, output length, latency, and cost per request. Freeze 20–30 hard prompts per workload as a regression set.

Days 6–8: caching check. Turn on caching diagnostics for each exact GPT route. Check that gpt-6.1-sol cached input is priced at $0.10 and that your prefixes remain stable enough to get the expected cache hits.

Days 9–12: staged cutover. Move 10%, then 50%, then 100% of traffic per workload, holding each step for at least a day. Keep the GPT-5.6 route warm as a one-line rollback.

Days 13–14: tier review. Check whether Terra workloads pass on GPT-6.1 Sol and whether some Sol workloads pass on GPT-6 Luna. The OpenAI tier guide covers the routing logic.

Finish before 21 November, when GPT-5.6 Sol's promotional price is scheduled to end.

When not to migrate yet

  • Regulated outputs with sign-off. If a workflow's outputs were validated by compliance or legal on GPT-5.6, budget for that review again. The price saving does not skip it.
  • Workloads that run in ChatGPT, not the API. Migrating there depends on OpenAI's rollout, not your code.
  • A frontier-model decision is pending. If you are also weighing Anthropic, test both at once. GPT-6 Sol vs Claude Opus 5.5 has the cost-per-task framing.

For a shortlist scored against your constraints, use the AI Model Picker. If the model sits inside a coding tool rather than your own code, the AI Coding Assistant Finder covers the harness side.

Sources (accessed 2026-09-26 unless noted)

Editorial note

AI Choice Engine publishes editorial guides to help readers understand fit, trade-offs, and next steps before choosing a tool or provider.

Newsletter

Get the buyer checklist that goes with this guide

Subscribe to download the matching checklist. Automated email delivery is still rolling out — this is not an inbox confirmation.

A practical scorecard for comparing fit, cost, rollout risk, support, and lock-in.

Updates only — checklists stay free from the resource library, with or without joining. Automated email delivery is still rolling out.

Next step

Use the live tool while the trade-offs are still fresh

The article gives context. The live tool turns those trade-offs into a clearer shortlist.

Buying guides

Guide pages connected to this article

These guides go one level deeper for readers who want a longer-form buying view before choosing a provider.

Keep reading

More articles in the same decision path

These pieces stay inside the same research journey instead of sending you somewhere unrelated.

  1. ChatGPT Free vs Plus in 2026 — Luna, Think, and the Sol sliderOpenAI's October plan page and Work/Codex model guide differ on GPT-6 Chat availability. Here's the dated August refresh, the current plan summary, and what to verify in your account.
  2. ChatGPT Work vs Chat vs Codex — which OpenAI surface are you buying?OpenAI's consumer plan page lists GPT-6 models for Plus and Pro, while its Work/Codex guide describes a different Chat availability picture. Keep surfaces and API billing distinct.

Next steps

Next step across the network

Continue with a focused hub page instead of restarting your research from scratch.