Best AI models for agents, by budget
The highest-ranked model in each price band, using only official list prices and public leaderboard scores. Verified 2026-11-04.
What it is A shortlist of the 14 models on our price list, grouped into three price bands.
Who it's for Builders picking a model for an AI agent who want a starting point at their budget.
How it compares The ranking follows public leaderboards, not our opinion, and every price links to its official source.
Under $1 per 1M tokens
| Model | Provider | Input /1M | Output /1M | Context | AA index | 1,000 calls* |
|---|---|---|---|---|---|---|
| Claude Haiku 5.5 · top pick | Anthropic | $0.100 | $0.500 | 1M | 36 | $0.450 |
| DeepSeek V4.1 Flash (1017) | DeepSeek | $0.300 | $1.20 | 1M | 33 | $1.20 |
| GPT-6 Luna | OpenAI | $0.100 | $0.500 | 1.05M | 28 | $0.450 |
$1 to $5 per 1M tokens
| Model | Provider | Input /1M | Output /1M | Context | AA index | 1,000 calls* |
|---|---|---|---|---|---|---|
| GPT-6.1 Sol · top pick | OpenAI | $2.00 | $10.00 | 1.05M | 55 | $9.00 |
| Claude Sonnet 5.5 | Anthropic | $2.00 | $10.00 | 1M | 55 | $9.00 |
| GPT-6.1 Terra | OpenAI | $1.00 | $5.00 | 1.05M | 51 | $4.50 |
| Gemini 4 Argon | $2.00 | $10.00 | 1M | 50 | $9.00 | |
| Grok 4.7 | xAI | $2.00 | $6.00 | 500K | 35 | $7.00 |
| Gemini 3.8 Flash | $0.750 | $3.75 | 1M | 31 | $3.38 | |
| DeepSeek V4 Pro (0813) | DeepSeek | $1.32 | $3.96 | 1M | 23 | $4.62 |
Over $5 per 1M tokens
| Model | Provider | Input /1M | Output /1M | Context | AA index | 1,000 calls* |
|---|---|---|---|---|---|---|
| Claude Opus 5.5 · top pick | Anthropic | $4.00 | $20.00 | 1M | 59 | $18.00 |
| GPT-6 Astra | OpenAI | $10.00 | $50.00 | 1.05M | 57 | $45.00 |
| Claude Fable 5.1 | Anthropic | $10.00 | $50.00 | 1M | 50 | $45.00 |
| Muse Wonder | Meta | $4.00 | $20.00 | 1.05M | 44 | $18.00 |
*Example workload: 1,000 calls of 2,000 input + 500 output tokens, at list price. Bands use blended price (3:1 input:output). Prices are USD per 1M tokens, standard tier list prices from official provider pages. Benchmarks not run by us; copied from public leaderboards. Omitted figures could not be verified.
Frequently asked questions
What is the best AI model for agents on a tight budget?
Among models with a blended list price under $1 per 1M tokens, Claude Haiku 5.5 ranks highest on our list (Artificial Analysis Intelligence Index 36). It costs $0.100 input and $0.500 output per 1M tokens.
How are these models ranked?
By Artificial Analysis Intelligence Index v5.0 (best published reasoning setting), then LMArena Text score. The scores are copied from public leaderboards; we don't run our own benchmarks. Your own tasks may rank models differently, so test on real prompts.
What does blended price mean?
Blended price weights input and output 3 to 1, a rough mix for agent calls that read more than they write. Each model page shows the separate input and output prices.
Do these prices include discounts?
No. They are standard list prices from each provider's official pricing page. Batch, cached-input and volume discounts are not included.