Ask for a model

Tell us the job. We’ll pick three.

Best overall, best on a budget and best regardless of price, with what each costs at your volume. No signup needed.

request.txt

I need help with , about , my budget is and what matters most is .

What we understoodCoding helpany budgetquality first10k a month
best_overall.txt
Best overall
88
  • Scores 100 on coding, the highest of 33 models we track for this job.
  • $42.00 per 1,000 requests.
  • Also the highest-scoring model for this job.
At 10k requests a month$420 a month
Passes our testProvisional score
best_budget.txt
Best budget pick
  • Scores 69 on coding, #14 of 33 models we track for this job.
  • $1.20 per 1,000 requests, 97% less than the best overall pick.
  • The cheapest model that passes our test for coding help for non-developers.
At 10k requests a month$12.00 a month
Passes our testOpen weightsProvisional score
RefineCheaperQuality firstMake it private

Also considered: Claude Sonnet 5.5 (match 78, $21.00 per 1,000), Claude Fable 5.1 (match 77, $105 per 1,000), Gemini 4 Argon (match 76, $21.00 per 1,000), DeepSeek V4.1 Flash 1017 (match 76, $2.70 per 1,000). 33 models were in the running.

alerts.txt
Model Menu will email you about this, plus the Friday letter. One click unsubscribes. We never sell or share your address.

How we pick

The ranking is a published formula, never a person’s choice and never paid for. We filter by the job, your budget and any hard needs such as open weights, then weigh quality and price by what you said matters most. Best overall is the top of that list. Best budget is the cheapest model that passes our test for the job. Best premium is the highest quality, whatever it costs.

Scores are provisional: they come from Arena and Epoch AI benchmarks until our own test rig runs. Prices are list prices, checked against OpenRouter where we can. Speed is not scored yet. See the methodology.