Frequently asked questions
Short answers to the questions we get most often about AI models and tooling.
All · 5
How is the price per 1M tokens calculated?
Providers bill input and output tokens separately. A token is roughly four characters of English text. For a typical chat request with a 2,000-token prompt and a 500-token answer, Claude 3.5 Sonnet costs about $0.0135 and GPT-4o about $0.01. Cached or batched requests are usually discounted.
Does a larger context window automatically mean better answers?
No. A larger window lets you include more material – a whole codebase or a long contract – but models still lose precision on details buried deep in very long inputs. For retrieval-heavy use cases, chunking plus a 128K window is often as good as stuffing 200K tokens into one prompt.
Which benchmark matters most for coding assistants?
HumanEval is the most quoted score, but it tests small isolated functions. SWE-bench Verified, which measures real GitHub issue resolution across a repository, correlates much better with day-to-day assistant quality. We list HumanEval for comparability and discuss SWE-bench in the guides.
Can I use the same prompts across Claude and GPT-4o?
Mostly. Both support system prompts, tool calling and JSON output. Claude responds well to XML-style structure, while GPT-4o has a dedicated structured-outputs mode. A thin routing layer such as the OpenAI-compatible endpoints or an SDK like LiteLLM lets you switch models without rewriting application code.
Are the figures on this site up to date?
We re-check published pricing and documentation monthly and note the date of the last check on every comparison. Vendors change prices and limits frequently; always confirm on the official pricing page before committing to a budget.