GPT 6.1 Sol: Near-Astra intelligence for a fifth of the price
TL;DR Highlight
GPT-6.1 Sol reportedly matches GPT-6 Astra at about one-fifth the cost.
Who Should Read
A backend developer choosing a coding-agent model by cost-performance across OpenAI and Anthropic.
Core Mechanics
- Sol 6.1 matches Astra at 1/5 cost
- A quoted announcement line says GPT-6.1 Sol matches GPT-6 Astra at roughly one-fifth of the cost.
- The article body was not provided, so detailed benchmarks can't be verified.
- Cached input got much cheaper
- Per a quoted line, cached input is $0.10 per million tokens.
- That is 95% less than standard input and 50% less than GPT-6 Sol's cached price.
- Higher effort didn't raise the score
- In the DeepSWE table a comment points to, GPT-6.1 Sol High scores 75.2% and XHigh 71.9%.
- The comment notes XHigh is also more expensive.
Community Reactions
- External test says it beats Opus 5.5
- A commenter ran 100 coding and engineering environments and found Astra and 6.1-Sol ahead of Opus 5.5, and cheaper via API.
- They say Anthropic leads only in chemistry, and GPT-6 models show low initiative beyond the prompt.
- Reports that GPT-6 Sol regressed
- One user switched to Opus 5.5 because GPT-6 Sol felt like a big regression from Sol 5.6.
- Another felt worse quality on simple refactoring versus 5.6-Sol.
- Opus slightly better on image-to-HTML, per one user
- A user ran the same flow and found Opus 5.5 better on alignment and animation polish.
- They said 6.1 Sol ran fast and cheap.
- Concern about rapid releases
- Some commenters say a new model days after GPT-6 Sol makes behavior shifts a liability.
- Plan changes noted
- A comment quoting Engadget says a $500 Pro plan was added and the $200 Pro allowance fell from 20x to 10x.
How to Apply
- Compare Sol 6.1 cost on your workload
- Run 5-10 real coding tasks on your current model and GPT-6.1 Sol today.
- Tabulate accuracy and token cost side by side.
- Calculate your cached-input share first
- Pull the cached input token ratio from last week's API logs.
- Plug in $0.10 per million tokens to estimate the monthly difference.
- Measure before raising effort
- Run the same tasks at High and XHigh several times.
- Compare average score and cost to see if the pricier setting actually helps.
Try It
# Paste into Claude or ChatGPT (simple model-comparison planning prompt)
I'm considering switching my coding-agent LLM. Candidates are my current model {{current model}} and GPT-6.1 Sol.
Assume cached input costs $0.10 per million tokens, monthly input tokens {{monthly input tokens}}, cache ratio {{cache ratio}}%.
1. Build a table of monthly cost differences.
2. Propose a checklist for evaluating 10 real tasks from our service.
3. Write an experiment procedure that repeats each setting 3 times and averages.
Next: paste your results and ask for cost-effectiveness by effort level.Terminology
Related Papers
Mercury 3: Diffusion at scale
Scaling diffusion LLM 40x gives Mercury 3: 4–8x faster than speed-tier models, similar quality
Timestral: Mistral's time-series foundation models
A 209M time-series model ranks first on three benchmarks and runs on CPU.
DeepSeek V4.1 Flash 1017 (Max)
DeepSeek V4.1 Flash 1017 Max scores 45 on the Intelligence Index at 211 tok/s and $0.30 input.
If you don't run your own LLM proxy, how do you know whats going on?
Running your own LLM proxy shows which keys, models, and cache misses drive your usage.
DeepSeek: DeepSeek v4.1 Flash 1017
DeepSeek V4.1 Flash 1017: $0.05/M input, 1M context, built for agents
DeepSeek-v4.1-Flash-1017
DeepSeek V4.1 Flash minor update lifts coding and agent benchmarks while cutting run time and output length
Show HN: Llmspeed – feel how fast a local LLM runs before you buy the hardware