ModelsGPT-6.1 Sol Pro: Pro Reasoning Mode, Explained and Costed
GPT-6.1 Sol Pro is the same model as GPT-6.1 Sol run in pro reasoning mode. How mode differs from effort, why the rate card hides the cost, and how to test it.
If you see GPT-6.1 Sol Pro in a model catalogue next to GPT-6.1 Sol with the identical price, the natural conclusion is that one of the two listings is wrong. Neither is. Sol Pro is not a separate model at all, and the identical per-token price is accurate. What differs is how many tokens a request consumes.
This guide explains what pro mode is, how it changes the bill, and how to decide whether a given workload should use it. For the base model's limits, pricing tiers and migration notes, read the GPT-6.1 Sol guide first.
Same weights, different execution path
OpenRouter describes openai/gpt-6.1-sol-pro as the same underlying model as GPT-6.1 Sol, served with reasoning.mode set to pro. The practical effect of the separate catalogue entry is that clients which only let you pick a model name, and cannot pass extra request fields, can still reach pro mode.
OpenAI documents the setting directly. GPT-5.6 and GPT-6 models support two reasoning modes in the Responses API, standard (the default) and pro, and OpenAI recommends pro for difficult tasks that need more model work and can tolerate higher latency and token usage.
Everything else is inherited from Sol: the 1,050,000-token context window, 128,000 tokens of output, text and image input, and the same tool set. If you call OpenAI directly, you do not need a different model name. You call gpt-6.1-sol and set the mode:
response = client.responses.create(
model="gpt-6.1-sol",
reasoning={"mode": "pro", "effort": "high"},
input="Find the race condition in this worker pool: ...",
)
Mode is not effort
This is the distinction most people miss. Sol already has five effort levels: low, medium, high, xhigh and max. Pro mode is a separate axis.
OpenAI's wording is that mode selects standard or pro execution, while effort controls how much reasoning the model applies within that mode. So standard at max and pro at medium are different configurations, and neither is strictly "more" than the other. You have a two-dimensional grid, not a longer slider.
The practical consequence: before concluding that a task needs pro mode, check whether raising effort in standard mode already gets you there. It is the cheaper experiment, and on many tasks it is enough. Reasoning models explained covers why more thinking helps some problems and does nothing for others.
How the bill actually changes
OpenAI says pro mode aggregates the model work performed to produce the final answer and bills those tokens at the selected model's standard rates. For Sol that means $2 per million input and $10 per million output, the same rate card as standard mode. OpenRouter lists Sol Pro at those rates too.
The per-token price is the same. The token count is not. OpenAI is explicit that pro mode performs more work than standard mode, increasing token usage and cost. OpenAI does not publish a typical multiplier, and it will vary by task, so measure it on your own prompts rather than assuming one.
Three implications follow:
- Rate-card comparisons mislead. A cost dashboard that multiplies list price by an assumed token count will undercount pro mode. Budget from measured usage.
- Output caps bite harder. Reasoning counts against
max_output_tokens, and a request that hits the cap returns incomplete after you have paid for the work. Leave more room in pro mode. - Latency rises. More work takes longer. Pro mode belongs on paths where nobody is watching a spinner.
Pro mode against the older Pro models
Earlier OpenAI Pro models were separate model IDs with their own premium price. GPT-5 Pro, for example, lists at $15 input and $120 output per million tokens, and OpenAI says existing Pro IDs keep that behaviour and pricing.
Pro mode on Sol changes the economics: you pay Sol's rates for every token, and the extra cost comes only from the extra work. Even if pro mode used several times the tokens of a standard request, the per-token gap against an old Pro model ID is large enough that Sol Pro is worth evaluating first for most new deep-reasoning work. That is an argument for testing it, not a claim that it matches any particular older model on quality.
A test plan for pro mode
Run the comparison on a fixed set of hard tasks, ideally ones a standard configuration has already failed:
- Sol, standard mode,
higheffort - Sol, standard mode,
maxeffort - Sol, pro mode, default effort
- GPT-6 Astra, standard mode, default effort
For each, record pass rate, total billed tokens and wall-clock time. Then work out cost per passed task. Pro mode earns its place when it fixes tasks that standard max could not, at lower cost per pass than moving up to GPT-6 Astra. If standard max matches it, use that and save the extra work.
A sensible production pattern is escalation: run standard Sol first, validate the result, and retry in pro mode only when validation fails. That keeps pro mode's token appetite confined to the requests that need it.
Common questions
Is GPT-6.1 Sol Pro a different model from GPT-6.1 Sol?
No. OpenRouter lists it as the same underlying model with reasoning.mode set to pro. Calling OpenAI directly, you use gpt-6.1-sol and set the reasoning mode to pro in the Responses API.
Why does GPT-6.1 Sol Pro cost the same per token as Sol?
OpenAI bills pro mode at the selected model's standard token rates. It costs more per request because it performs more model work, so it consumes more tokens, not because the rate is higher.
Should I use pro mode or raise reasoning effort?
Try raising effort in standard mode first, since it is the cheaper experiment. Mode and effort are separate settings, so if standard max still fails your hard tasks, test pro mode next and compare cost per passed task.
