How much cheaper are these than frontier models?

Typically 10 to 50 times cheaper per token; exact prices render beside each pick from the live catalog. The in-app estimate shows the cost of a message before you send it.

What is the free option?

The idapt free tier includes a daily usage allowance on the affordable and standard model bands, no card required, and local models via Ollama route free through your own machine.

All roundups

Best of

Frontier answers at commodity prices.

The models that deliver the most per dollar, with live per-million-token prices from the catalog.

Last reviewed July 16, 2026

Try idapt free Browse all models

Most prompts do not need a frontier model. The picks below sit at the steep part of the price-performance curve: strong general capability at prices where you stop thinking about cost per request. Prices render live from the catalog, per million tokens, so this page cannot quietly go stale.

How we pickedPicks minimize price for acceptable benchmark standing rather than maximizing raw scores: each one holds its own on the general lanes of the live rankings while pricing an order of magnitude under the frontier. Prices and context beside each pick come from the live registry.

The picks

1Gemini 3.5 FlashThe default fast model.

$1.50 in / $9 out per M tokens · 1.0M context

Google's speed line keeps inheriting the Pro line's capabilities a tier down in price; multimodal, long context, and quick.

Best for: The everyday default for routine prompts.

2DeepSeek V4 FlashThe price floor with real capability.

$0.09 in / $0.18 out per M tokens · 1M context

Among the cheapest ways to run a capable model, with the V4 family's strength intact for summaries, extraction, and drafting.

Best for: High-volume pipelines and background jobs.

3Grok 4.1 FastThe low-latency conversationalist.

Tuned for snappy interactive use; the model that makes voice mode and quick back-and-forth feel instant.

Best for: Interactive chat and voice where latency is the feature.

4Claude Haiku 4.5The cheap model with judgment.

$1 in / $5 out per M tokens · 200K context

Anthropic's small model keeps an unusual share of the family's instruction-following and tone, which matters when cheap output still faces readers.

Best for: User-facing text at scale.

5Mistral Small 3The efficient European open-weight.

$0.15 in / $0.6 out per M tokens · 262K context

Open weights, efficient serving, and solid multilingual performance; also small enough to run locally.

Best for: Multilingual work and local deployment.

Frequently asked

Run them in one workspace

Multi-Model Chat

GPT, Claude, Gemini, Grok, Llama: pick mid-conversation, cheapest-first.

Local AI

Run open models on your own hardware with Ollama: private, free, no rate limits.

Keep comparing

Serious models, your hardware.Open weights at the frontier.Models that show their work.The models that cost nothing to run.

Judge them on your own prompt

Every pick on this page runs in one idapt chat. Start free with a daily allowance, no credit card.

Try idapt free