Preview build — figures marked “✓” or in green are sourced (benchmarks by Artificial Analysis); treat everything else (value scores, “Est.” figures, unmarked prices and changes) as illustrative. Methodology
Cost per task
What a finished task actually costs.
Per-token, per-second, per-character and per-image prices can't be compared directly — finished tasks can. Each category below defines one standard task with its conversion published. Quality is shown per model but never compared across categories — a video Elo and a language-model index measure different things.
$1 buys
757×
summarize a report · Gemini 2.5 Flash-Lite
$1 buys
1.9×
a 10-second clip · Seedance 1.5
$1 buys
277×
one minute of narration · Cloud TTS WaveNet
$1 buys
333×
one image · FLUX Schnell
Summarize a report
LLMA 1,000-word summary of a ~6,000-word document in one request. Conversion: ≈8,000 input + 1,300 output tokens (1.33 tokens per English word).
| Model | Provider | Cost / task | Tasks / $1 | Intelligence index |
|---|---|---|---|---|
| OpenAI | $0.0013 | 757 | — | |
Gemini 2.5 Flash-Litefrontier | $0.0013 | 757 | 6.9 | |
Qwen 3 235Bfrontier | Together | $0.0024 | 420 | 18.2 |
| OpenAI | $0.0053 | 189 | — | |
| $0.0056 | 176 | — | ||
| Mistral | $0.0060 | 168 | 15.9 | |
| Anthropic | $0.015 | 68 | — | |
| OpenAI | $0.015 | 68 | — | |
Gemini 2.5 Profrontier | $0.023 | 43 | 25.8 | |
| OpenAI | $0.026 | 37 | 19.4 | |
o3frontier | OpenAI | $0.026 | 37 | 30.4 |
Claude Sonnet 5frontier | Anthropic | $0.029 | 34 | 53.4 |
| $0.032 | 31 | — | ||
| OpenAI | $0.032 | 31 | 42.2 | |
| Anthropic | $0.072 | 13 | 34.7 |
A 10-second clip
VideoTen seconds of generated video at each model's priced tier. Conversion: 10 × the per-second list price (tier caveats per row on the pricing page).
| Model | Provider | Cost / task | Tasks / $1 | Arena Elo |
|---|---|---|---|---|
Seedance 1.5frontier | fal.ai | $0.520 | 1.9 | 1172 ±8 |
Kling 2.5frontier | fal.ai | $0.700 | 1.4 | 1200 ±7 |
Wan 2.7frontier | fal.ai | $1.00 | 1.0 | 1216 ±9 |
| fal.ai | $1.00 | 1.0 | 1186 ±8 | |
| $1.00 | 1.0 | 1201 ±8 | ||
| Replicate | $1.68 | 0.6 | 1212 ±8 | |
Kling 3.0 Profrontier | Replicate | $2.24 | 0.4 | 1240 ±8 |
Seedance 2.0frontier | fal.ai | $3.03 | 0.3 | 1266 ±8 |
| $4.00 | 0.3 | 1201 ±8 |
One minute of narration
TTSSixty seconds of spoken audio at a typical narration pace. Conversion: ≈150 words/min ≈ 900 characters, so 0.9 × the per-1K-chars price.
| Model | Provider | Cost / task | Tasks / $1 | Arena Elo |
|---|---|---|---|---|
| $0.0036 | 277 | — | ||
Cloud TTS WaveNetfrontier | $0.0036 | 277 | 902 ±12 | |
tts-1frontier | OpenAI | $0.013 | 74 | 1083 ±11 |
| Amazon | $0.014 | 69 | 879 ±14 | |
tts-1-hdfrontier | OpenAI | $0.027 | 37 | 1102 ±12 |
| ElevenLabs | $0.045 | 22 | 1098 ±11 |
One image
ImageA single generated image at the model's priced resolution. Conversion: 1 × the per-image list price.
| Model | Provider | Cost / task | Tasks / $1 | Arena Elo |
|---|---|---|---|---|
FLUX Schnellfrontier | fal.ai | $0.0030 | 333 | 1000 ±0 |
| OpenAI | $0.040 | 25 | 950 ±8 | |
Imagen 4.0frontier | $0.040 | 25 | 1101 ±8 | |
| fal.ai | $0.040 | 25 | 1088 ±7 | |
GPT Image 2frontier | OpenAI | $0.053 | 18 | 1339 ±7 |
| $0.067 | 14 | 1262 ±7 | ||
| fal.ai | $0.068 | 14 | 1240 ±9 | |
| $0.134 | 7.5 | 1225 ±7 |
Costs are computed from the sourced list prices shown on the pricing pages (excluding caching, batch discounts, volume and context-length tiers). Conversion constants are editorial and versioned with the methodology. Benchmarks by Artificial Analysis; “—” means no sourced measurement exists. See methodology.