Batch API AI models: about half price if the answer can wait
| # | AI model | Capabilities | API $/1M (in / out) | Context window | Price / mo | Output $/1M | Overview | Visit site |
|---|---|---|---|---|---|---|---|---|
| 1 |  Claude Fable 5.1 Anthropic |    | $10 / $50 | 1M tokens | API only | 50 | View | |
| 2 |  GPT-6 Astra OpenAI | | $10 / $50 | 1.05M tokens | API only | 50 | View | |
| 3 | GPT-6.1 Sol OpenAI | | $2 / $10 | 1.05M tokens | API only | 10 | View | |
| 4 | GPT-6 Sol OpenAI | | $2 / $10 | 1.05M tokens | API only | 10 | View | |
| 5 | GPT-6 Luna OpenAI | | $0.1 / $0.5 | 1.05M tokens | API only | 0.5 | View | |
| 6 | Claude Opus 5.5 Anthropic | | $4 / $20 | 1M tokens | API only | 20 | View | |
| 7 | Claude Sonnet 5.5 Anthropic | | $2 / $10 | 1M tokens | API only | 10 | View | |
| 8 | Claude Opus 5 Anthropic | | $5 / $25 | 1M tokens | API only | 25 | View | |
| 9 | Claude Fable 5 Anthropic | | $10 / $50 | 1M tokens | API only | 50 | View | |
| 10 | GPT-5.6 Sol OpenAI | | $4 / $20 | 1.05M tokens | API only | 20 | View | |
| 11 |  GLM 5.3 Z.ai |  | $1.4 / $4.4 | 1.31M tokens | API only | 4.4 | View | |
| 12 |  Grok 4.6 xAI | | $2 / $6 | 500K tokens | API only | 6 | View | |
| 13 |  Kimi K3 Moonshot AI |  | $3 / $15 | 1.05M tokens | API only | 15 | View | |
| 14 | GPT-5.6 Terra OpenAI | | $2 / $12 | 1.05M tokens | API only | 12 | View | |
| 15 | Claude Opus 4.8 Anthropic | | $5 / $25 | 1M tokens | API only | 25 | View | |
| 16 | GLM 5.3 Flash Z.ai | | $0.075 / $0.25 | 1.31M tokens | API only | 0.25 | View | |
| 17 |  Gemini 3.8 Flash Google |  | $0.75 / $3.75 | 1.05M tokens | API only | 3.75 | View | |
| 18 |  Qwen3.8 Max (0902) Qwen | | $2 / $6 | 1M tokens | API only | 6 | View | |
| 19 | Qwen3.8 2.4T A95B Qwen |  | $2 / $6 | 1.05M tokens | API only | 6 | View | |
| 20 | Gemini 3.7 Flash Google | | $0.75 / $3.75 | 1.05M tokens | API only | 3.75 | View |
Browse by need
Batch processing queues your requests for asynchronous execution and charges roughly half the standard rate. There is no latency guarantee - results arrive within hours rather than seconds - which makes it the single largest lever on cost for any workload that does not need an immediate answer.
140 models offer it. Typical candidates: overnight document pipelines, bulk classification, dataset labelling, embedding backfills, scheduled reports. Anything driving a live chat window is not a candidate. Combined with prompt caching, a batch pipeline on a mid-tier model lands close to budget-tier pricing while keeping mid-tier quality.
Models with this capability
The 140 models below all support this capability. Prices are input and output per 1M tokens from the vendor official price list; for dirhams multiply by 3.6725. Tap a card to open the model page.
How to choose
- Identify which workloads genuinely tolerate delay — most back-office work does.
- Size batches to the vendor limits rather than sending one giant job.
- Build retry handling: partial failures inside a batch are normal.
- Recheck economics after enabling caching; the combination often changes the model choice.
Common mistakes
Who this suits
Related filters
Frequently asked questions
How long does a batch take?
What workloads suit batch?
Can I combine batch with caching?
Is quality identical to standard mode?
Can I cancel a running batch?
Frequently searched
Rate this page
Help us improve