Sol 6.1 vs Sol 6: API prices, with Astra for comparison
Sol 6.1 and Sol 6 both cost $2 per million uncached input tokens and $10 per million output tokens. Sol 6.1 cuts cache reads from $0.20 to $0.10. Astra charges $10 input and $50 output.
Standard API prices per million tokens
USD, for requests with at most 272,000 input tokens. Each rate below links to its official model documentation.
| Model | Uncached input | Cache reads | Cache writes | Output | Source |
|---|---|---|---|---|---|
| GPT-6.1 Sol | $2.00 | $0.10 | $2.50 | $10.00 | OpenAI ↗ |
| GPT-6 Sol | $2.00 | $0.20 | $2.50 | $10.00 | OpenAI ↗ |
| GPT-6 Astra | $10.00 | $1.00 | $12.50 | $50.00 | OpenAI ↗ |
Compare your Sol 6.1, Sol 6 and Astra bill
Use the same token counts for all three models. Cache percentages divide your input into reads, writes and uncached tokens.
Standard rates: input does not exceed 272,000 tokens per request.
| Model | Uncached input | Cache reads | Cache writes | Output | Monthly total |
|---|---|---|---|---|---|
| GPT-6.1 Sol | $120.00 | $24.00 | $0.00 | $500.00 | $644.00 |
| GPT-6 Sol | $120.00 | $48.00 | $0.00 | $500.00 | $668.00 |
| GPT-6 Astra | $600.00 | $240.00 | $0.00 | $2,500.00 | $3,340.00 |
Sol 6.1 saves $24/month versus Sol 6 for the example above.
Output includes billable reasoning tokens. Cache-write tokens replace uncached input in this estimate. Tool calls, image tokens, regional premiums, taxes and negotiated discounts are excluded. Fast is unavailable with Sol 6.1 EU data residency. Check eligibility before choosing a processing tier.
What “GPT 6 vs 6.1” means for this cost comparison
GPT-6 includes distinct API models. We compare gpt-6.1-sol with its earlier Sol version, gpt-6-sol, and separately show gpt-6-astra. Keeping the API IDs explicit matters: substituting Astra for Sol changes the uncached input and output rates by a factor of five.
The Sol 6.1 prices above are for the OpenAI API. The release announcement also lists Codex and ChatGPT Work access. A search for “Codex Sol 6.1” may lead to a plan with included usage and limits. A subscription price does not turn into a per-token API rate, so check the billing surface you use before estimating a monthly budget.
Worked example: repeated context in an agent
Suppose an application makes 100,000 requests a month. Each request sends 3,000 input tokens, 80% of which are already cached, and generates 500 output tokens. Across the month, that is 60 million uncached input tokens, 240 million cache-read tokens and 50 million output tokens.
- Sol 6.1: $120 input + $24 cache reads + $500 output = $644/month.
- Sol 6: $120 input + $48 cache reads + $500 output = $668/month.
- Astra: $600 input + $240 cache reads + $2,500 output = $3,340/month.
Sol 6.1 saves $24, or about 3.59% of the Sol 6 bill in this example. The cache-read component falls by half; output still accounts for most of the bill. Increasing the reused context or decreasing output length makes the cache price difference more visible.
Cache writes and long prompts can change the result
A cache read uses context that is already available. A cache write creates cached context and has a different rate. In the calculator, reads, writes and uncached input are separate shares of the same input total. If a request sends 10,000 input tokens with 70% reads and 10% writes, the remaining 20% uses the uncached rate.
Cache hits are an assumption to measure against actual API usage. A repeated prompt does not guarantee a hit. Start with the cache-read tokens reported by your API calls, and include the writes needed to create or refresh that cache.
Once a request exceeds 272,000 input tokens, all three models apply 2× input and cache rates and 1.5× output rates to the full request. The threshold counts all input, including cached tokens. For Sol 6.1 that means $4 uncached input, $0.20 cache reads, $5 cache writes and $15 output per million tokens.
For example, one Standard request with 300,000 uncached input tokens and 10,000 output tokens costs $1.35 on either Sol version and $6.75 on Astra. Charging the higher rate only on the 28,000 input tokens above the threshold would understate the bill.
Batch, Flex and Fast pricing
Batch and Flex cost 50% of the applicable Standard rates. Fast costs 2×. The processing choice in the calculator applies after the long-context adjustment. For the $644 Sol 6.1 example, that produces $322 with Batch/Flex or $1,288 with Fast, assuming the same token usage and cache shares.
Processing modes have different availability and latency constraints. In particular, Sol 6.1 Fast is unavailable with EU data residency. The model documentation also lists a 10% regional-processing premium where available; that premium is excluded from this calculator.
Choosing a model from the price comparison
For equal token counts, Sol 6.1 costs the same as Sol 6 with uncached input and costs less when there are cache reads. Its Standard input and output rates are one fifth of Astra's. Those ratios compare unit prices; they do not establish equal quality, equal latency or equal tokens needed to finish a task.
To compare cost per completed task, run a representative sample on each model. Record successful outcomes, retries, total input, cache reads and billable output, including reasoning tokens. A task that needs more attempts or longer reasoning can change the apparent saving. Tool charges and other services can also add to the token bill.
Use the estimate above for this three-model decision, the full API cost calculator for other providers, or the OpenAI price table for the wider catalog.
Official sources
Prices and billing rules checked on 2026-10-01: GPT-6.1 Sol documentation · GPT-6 Sol documentation · GPT-6 Astra documentation. API access, processing eligibility and list prices can change; these links let you check the same numbers before committing to volume.