OpenAI released GPT-6.1 Sol on September 29, one week after GPT-6 Sol and GPT-6 Luna filled in the two tiers below the GPT-6 Astra flagship. It keeps GPT-6 Sol's $2 and $10 per million input and output tokens and cuts cached input to $0.10.
That makes the Sol tier much harder to overlook. On independent benchmarks, 6.1 Sol scores four points above GPT-6 Sol and one point below Astra at max effort, for about a fifth of Astra's cost per task.
The tier names hide how much the lineup overlaps. From high effort up, 6.1 Sol matches Astra one effort level down for a fraction of the cost, while Luna still covers high-volume work for under $0.10 per task. With three tiers and five or six effort levels each, the real choice is a tier and an effort setting for every step of an agent.
So the question with GPT-6.1 Sol vs Luna vs Astra is which one to use for each step in 2026.
What are GPT-6.1 Sol, Luna, and Astra?
GPT-6 Astra is the flagship of the current GPT-6 family, GPT-6.1 Sol is the middle tier, and GPT-6 Luna is the entry tier. All three share a 1.05M-token context window and a 128K-token output limit, and OpenAI says Luna was trained with methods similar to Astra's.
GPT-6.1 Sol
GPT-6.1 Sol is OpenAI's upgrade to GPT-6 Sol, released a week after it, though OpenAI hasn't deprecated the older model and gpt-6-sol is still available in the API. It costs $2 per million input tokens, $0.10 per million cached input tokens, and $10 per million output tokens, and its reasoning effort runs from low to max, with medium as the default and no none setting.
The name is where it gets confusing. In GPT-5.6, Sol was the frontier tier, with Terra in the middle and Luna at the bottom. GPT-6 puts Astra above Sol, so the same name now sits one rung lower, and gpt-6.1-sol is a separate model from both gpt-6-sol and gpt-5.6-sol.
GPT-6 Luna
GPT-6 Luna is the GPT-6 model OpenAI builds for focused, high-volume tasks, where each call is short and the output is easy to check.
At $0.10 input and $0.50 output per million tokens, Luna costs one twentieth of 6.1 Sol and one hundredth of Astra on both lines. Its reasoning effort runs from none to max, which makes it the only one of the three with a none setting.
GPT-6 Astra
GPT-6 Astra sits above both as the model OpenAI positions for the hardest end-to-end work. At $10/$50 per million tokens it costs five times 6.1 Sol's price, and its effort range runs from low to max, the same as 6.1 Sol.
GPT-6.1 Sol vs Luna vs Astra: overview
All three GPT-6 models accept text and images and return text, with the same context window and output limit. The differences are in price and effort range.
| GPT-6 Astra | GPT-6.1 Sol | GPT-6 Luna | |
|---|---|---|---|
| API model ID | gpt-6-astra | gpt-6.1-sol | gpt-6-luna |
| Input / output per 1M | $10 / $50 | $2 / $10 | $0.10 / $0.50 |
| Cached input per 1M | $1.00 | $0.10 | $0.01 |
| Context window | 1.05M tokens | 1.05M tokens | 1.05M tokens |
| Max output | 128K tokens | 128K tokens | 128K tokens |
| Reasoning effort | low to max | low to max | none to max |
| Knowledge cutoff | Apr 30, 2026 | Apr 30, 2026 | May 18, 2026 |
| Best for | Fast answers on hard steps | Default for most agent steps | High-volume, checkable steps |
Because the spec sheets barely differ, choosing between the three comes down to what each tier delivers per dollar at a given effort level.
GPT-6.1 Sol vs Luna vs Astra: pricing
Output tokens cost five times the input rate on all three GPT-6 tiers, and reasoning tokens bill as output. As a result, the effort level you pick often moves the bill more than the tier's headline input price.
Per-token pricing
On standard input and output, Astra costs five times 6.1 Sol, and 6.1 Sol costs twenty times Luna. Cached input breaks that pattern, since 6.1 Sol's $0.10 rate is a tenth of Astra's and ten times Luna's. Compared with GPT-5.6, 6.1 Sol and Luna come in at exactly half of the previous promotional rates of $4/$20 and $0.20/$1.20.
The service-tier multipliers are the same across the family. Batch and Flex are half of Standard, Fast mode is double, and regional processing adds 10% where it's available. EU data residency narrows those options differently by model: 6.1 Sol can't use Fast mode with EU residency, and Luna only supports Standard processing with it, so an EU Luna workload can't pair residency with the Batch or Flex discount.
Caching and long-context pricing
Cached input bills at 10% of the uncached rate on Astra and Luna and at 5% on 6.1 Sol, and cache writes bill at 1.25x on all three. Caching also shapes routing by step, because a cached prefix belongs to the model that processed it, so moving a conversation from 6.1 Sol to Astra starts the cache over.
Long prompts have their own threshold. Once input passes 272K tokens, the whole request bills at 2x the input and cache rates and 1.5x the output rate, so a 300K-token 6.1 Sol request pays $1.20 for input alone where a 270K-token request pays $0.54. If most of that context repeats between calls, prompt caching is the first place to recover it.
Cost per task
Per-token prices don't account for how many tokens each model spends. On the Artificial Analysis Intelligence Index at max effort, 6.1 Sol uses about 38K output tokens per task against Astra's 27K.
Because 6.1 Sol spends more tokens, the per-task gap is a little narrower than the rate card suggests. Astra at max costs $3.26 per task against 6.1 Sol's $0.72, about four and a half times as much where the per-token ratio is five, while Luna at max costs $0.07.
Effort level stretches the range further. 6.1 Sol costs $0.13 per task at low and $0.72 at max, a spread of more than five times inside one model.
GPT-6.1 Sol vs Luna vs Astra: performance
6.1 Sol is the model that moved in this lineup, scoring four points above GPT-6 Sol at max effort, while Luna still scores close to its GPT-5.6 predecessor. Astra still scores highest at every effort level the three tiers share, but only by one to four points over 6.1 Sol.
Intelligence and coding
On the same Intelligence Index, a composite of 10 evaluations, Astra scores 53 at max effort, 6.1 Sol 52, and Luna 37. Artificial Analysis's launch analysis put 6.1 Sol four points above GPT-6 Sol and five above GPT-5.6 Sol, with gains across its knowledge work and coding evaluations.
Coding moved the same way. On the Coding Agent Index, run in OpenAI's Codex harness, 6.1 Sol at max scores 60, three points above GPT-6 Sol and two below Astra, while Luna scores 41, down two from its GPT-5.6 predecessor.
Knowledge work
Knowledge work is where GPT-6 Sol had given ground to GPT-5.6 Sol, and 6.1 Sol wins back some of it. On GDPval-AA, a benchmark adapted from OpenAI's dataset of occupational tasks, 6.1 Sol gains 5 points over GPT-6 Sol, and it adds 4 on AA-Briefcase, a long-horizon knowledge work evaluation. Its hallucination rate at max effort also fell from 60% to 54%, while accuracy rose 8 points to 62%.
Luna hasn't had the same update. Against GPT-5.6 Luna, it dropped about 75 Elo on GDPval-AA and about 45 on AA-Briefcase, and its hallucination rate sits at 77%. For a step whose output someone reads and acts on, such as a migration plan or a written report, that's a reason to score Luna against your own rubric before moving the step over.
Speed and latency
Across effort levels, 6.1 Sol generates 59 to 66 output tokens per second against Astra's 44 to 51. Speed figures for Luna aren't available yet.
Effort level drives time to first token far more than tier does. Astra returns its first token in about 3 seconds at low and 6.2 at medium, then jumps to 58 seconds at high and 315 at max. 6.1 Sol follows nearly the same curve, from 2.1 seconds at low and 5.7 at medium to 58 at high and 273 at max, though at xhigh it takes about 109 seconds against Astra's 202.
For any step a user waits on, low or medium effort on either tier keeps the delay in single-digit seconds.
Performance by effort level
The same benchmarks score each model at every effort level it supports, including GPT-6 Luna, and 6.1 Sol overlaps with Astra at almost every level.
| Effort | GPT-6 Astra | GPT-6.1 Sol | GPT-6 Luna |
|---|---|---|---|
| max | 53 · $3.26 | 52 · $0.72 | 37 · $0.07 |
| xhigh | 52 · $2.31 | 51 · $0.39 | 34 · $0.04 |
| high | 51 · $1.73 | 50 · $0.32 | 32 · $0.03 |
| medium | 50 · $1.54 | 48 · $0.21 | 29 · $0.02 |
| low | 46 · $0.82 | 42 · $0.13 | 21 · $0.0045 |
| none | n/a | n/a | 18 · $0.01 |
Intelligence Index score · cost per task
From high effort up, 6.1 Sol matches Astra one level down. Sol at max ties Astra at xhigh (52) for $0.72 against $2.31, and Sol at high ties Astra at medium (50) for $0.32 against $1.54. Sol at medium (48) also outscores Astra at low (46) for about a quarter of the cost per task.
What you give up is latency at the top. Sol at high takes about 58 seconds to its first token, where Astra at medium reaches the same score and starts answering in about 6.
Luna's none setting also costs more than it looks. Luna with no reasoning scores 18 and costs $0.01 per task, more than Luna at low, because it spends more output tokens per task. That matters if you call Luna through Chat Completions, which only supports function calling at none, while 6.1 Sol can't make tool calls through Chat Completions at any effort.
Which is best: Sol, Luna, or Astra?
No single GPT-6 model is the right choice for a whole agent. Astra scores highest at every effort level it shares with 6.1 Sol, but by only one or two points from medium up, and 6.1 Sol reaches each of those scores for about a fifth of the cost, so the useful question is which tier and effort to use for each kind of step.
Pick a tier and an effort level
Across the benchmark data, a handful of configurations cover most agent steps:
- GPT-6.1 Sol at medium - The default for general agent work, tool calling, and code edits. It scores 48 for $0.21 per task and returns its first token in under 6 seconds.
- GPT-6.1 Sol at high or xhigh - For hard steps nobody is waiting on, like planning a multi-file change in a background job. It scores 50 to 51 for $0.32 to $0.39 per task, matching Astra at medium and high for about a fifth of the cost.
- GPT-6 Astra at low or medium - For a hard step that also needs a fast first token, like recovering after a tool call fails mid-conversation. Astra at medium scores 50 and starts answering in about 6 seconds, where 6.1 Sol needs high effort and close to a minute for the same score.
- GPT-6 Astra at max - For the few steps where one more point of accuracy matters more than cost. It scores 53 for $3.26 per task.
- GPT-6 Luna at high or max - For extraction, classification, and routing, where the output has a right answer you can check. Luna at max scores 37 for $0.07 per task, a third of Claude Haiku 5.5's cost per task at max effort, which is the core tradeoff in any Haiku vs Luna decision.
The setting to question is 6.1 Sol at max. It adds one point over Sol at xhigh for nearly twice the cost per task and takes about four and a half minutes to its first token, so xhigh is usually the ceiling for Sol.
Route each agent step to a different tier
Every call in an agent run can set its own model and effort, so the tier decision can happen per step instead of once per application. A coding agent might triage an incoming issue with Luna at high, plan the change with 6.1 Sol at xhigh, make the edits with 6.1 Sol at medium, and write the pull request summary with Luna again. If a user is waiting on the plan, Astra at medium gets a comparable score back in seconds.
Where those boundaries go depends partly on caching. Switching models restarts the prompt cache, so the split works best at natural handoffs, such as passing a finished plan to the step that executes it. Tool calling adds one more constraint, since every tool call on 6.1 Sol, and any tool call on Luna above none, needs the Responses API.
A fixed split like this is simpler to reason about than an LLM router that picks a model per request from its content. Each step's assignment is also a single setting you can score on your own traffic, with a rubric-based evaluator such as LLM-as-a-judge.
Route GPT-6.1 Sol, Luna, and Astra with Respan
Routing agent steps across GPT-6 tiers pays off when switching models takes one change and every call's cost is visible afterward. Respan lets you route, observe, and evaluate every LLM call, and GPT-6 Astra, GPT-6.1 Sol, and GPT-6 Luna are 20% off on Respan during the launch deal.
Respan's LLM gateway puts GPT-6.1 Sol, GPT-6 Luna, and Astra behind one OpenAI-compatible endpoint with 1,000+ models, so moving a step to a different tier means changing one word in the request.
- Fallback when a provider fails - Define a fallback chain once, at the org level or per request, and traffic moves to the next model in the chain when a provider errors or rate-limits.
- Cost for every step - Each request is logged as a span with latency and cost attached, and spend breaks down by model, request, and end customer, with hard limits that stop spend at a ceiling.
- Traces across the whole run - Every LLM call, tool run, and agent turn nests in one trace, so a bad output leads back to the step and model that produced it.
- Evals before and after a switch - Build datasets from production logs, run them through each model with your evaluators, then deploy the same evaluators on live traffic so a regression after moving a step to Luna surfaces in real time.
FAQ
Is GPT-6.1 Sol the same as GPT-6 Sol?
No. gpt-6.1-sol is a separate model from gpt-6-sol, with the same $2/$10 per million tokens but cached input at $0.10 instead of $0.20. It drops the none effort setting, running from low to max, and scores four points higher on independent benchmarks at max effort.
Is there a GPT-6 Terra?
OpenAI hasn't released a GPT-6 Terra. The GPT-6 family is Astra, Sol, and Luna, and GPT-5.6 Terra is still listed in OpenAI's model catalog as the option that balances intelligence and cost.
Which ChatGPT plans include GPT-6.1 Sol and Luna?
GPT-6.1 Sol and GPT-6 Luna are in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users. Free and Go users can use Luna in the desktop app, and neither model is in Chat yet.
Does GPT-6.1 Sol support tool calling in Chat Completions?
No. OpenAI's GPT-6.1 Sol model page supports Chat Completions only without tool calling, so any tool call needs the Responses API. GPT-6 Luna supports tool calls in Chat Completions only at none effort, and Astra's model page doesn't list the same restriction.
Can I use GPT-6.1 Sol, Luna, and Astra in the same app?
Yes. Through the Respan gateway, all three sit behind one endpoint, so each call can name its own GPT-6 model and effort level, and every call's cost and trace land in one place. OpenAI's API also accepts any of the three on each request, since they share the same endpoints.
