Back to all plans
百川 API
Baichuan4-Air
Baichuan open API — Baichuan4 family uses combined billing, M series splits input/output, billed per 1K tokens from account balance
Token API
Core models
Baichuan-M3-PlusBaichuan-M3Baichuan-M2-PlusBaichuan-M2Baichuan4-TurboBaichuan4-AirBaichuan4Baichuan3-TurboBaichuan3-Turbo-128kBaichuan2-TurboBaichuan2-53BBaichuan-Text-Embedding
Baichuan-M3-Plus
Medical vertical M3 Plus—¥5 input, ¥9 output per 1M tokens (32k); auto medical search ¥0.03/call.
Baichuan-M3
M3 general reasoning—¥10 input, ¥30 output per 1M tokens (32k), split billing.
Baichuan-M2-Plus
M2 Plus—same as M3 pricing (¥10/¥30 per 1M); auto medical search.
Baichuan-M2
M2 light tier—¥2 input, ¥20 output per 1M tokens (32k), lowest M-series input price.
Baichuan4-Turbo
Baichuan4 Turbo—combined ¥15 per 1M tokens (32k), mid Baichuan4 tier.
See the official site for more models
Additional core model names still appear above, with full details on the latest official page.
Plan details
Baichuan4-Air
Lowest combinedRecommendedInput
¥0.98 per 1M tokens
Output
Combined input+output · 32k context
Baichuan4-Air
Lowest combinedRecommendedInput
¥0.98 per 1M tokens
Output
Combined input+output · 32k context
Usage
Baichuan4-Air is the lowest-priced general model on the pricing page—combined input+output billing at ¥0.98 per 1M tokens, 32k context—for large-scale daily chat and budget-sensitive workloads.
Models
Combined billing means prompt and completion are not priced separately—read bills as total token usage × unit price, unlike the split-billing M series.
Highlights
Suited to support, lightweight assistants, routine Q&A, and steady traffic needing minimal per-call cost; pass with_search_enhance=false to skip ¥0.03/search when search is not needed.
Best for
Teams running low-cost production traffic, routine chat, and budget-sensitive services
Baichuan-M3-Plus
Medical verticalInput
¥5
Output
¥9
Usage
Baichuan-M3-Plus bills input and output separately: ¥5/M input, ¥9/M output, 32k context—covering all tokens from the full conversation flow.
Models
Calls auto-trigger medical search billed at ¥0.03 per call—factor search volume on top of model tokens for true cost.
Highlights
Suited to healthcare, professional consulting, and other verticals needing M-series reasoning plus acceptable search add-on costs.
Best for
Healthcare verticals, professional consulting, and teams needing M3-Plus capability
Usage
Baichuan-M2 at ¥2/M input and ¥20/M output with 32k context is the lowest input-priced split-billing model in the M family.
Models
Output is priced higher than input—best when prompts are long and completions short; if output-heavy, compare with combined Baichuan4-Turbo at ¥15/M.
Highlights
Suited to lightweight reasoning and tool use needing M-series capability with input-heavy, prompt-cost-sensitive workloads.
Best for
Input-heavy workloads needing M2 split billing for lightweight reasoning
Baichuan4-Turbo
Baichuan4 TurboInput
¥15 per 1M tokens
Output
Combined input+output · 32k
Usage
Baichuan4-Turbo bills combined at ¥15 per 1M tokens with 32k context—mid-tier between Air (¥0.98) and Baichuan4 flagship (¥100).
Models
Combined billing avoids separate prompt/completion estimates—suited to production tasks needing more than Air without the ¥100 tier.
Highlights
Suited to general chat, content generation, and teams needing Baichuan4 capability on a tighter budget.
Best for
General production needing Baichuan4 capability at moderate budget
Usage
Baichuan-M3 at ¥10/M input and ¥30/M output with 32k context; M2-Plus matches pricing but auto-triggers medical search.
Models
General M-series reasoning tier—higher unit price than M3-Plus but no mandatory medical search—for general agents and multi-turn chat.
Highlights
If output-heavy, compare output ¥30/M against combined Baichuan4 at ¥100/M total tokens holistically.
Best for
General reasoning and agent workloads needing M-series split billing
Baichuan4
Baichuan4 flagshipInput
¥100 per 1M tokens
Output
Combined input+output · 32k
Usage
Baichuan4 bills combined at ¥100 per 1M tokens with 32k context—the top Baichuan4 tier on the pricing page for quality-sensitive professional tasks.
Models
Unit price is far above Air and Turbo—use as a key-path model, not the default for all traffic.
Highlights
Suited to complex content generation, deep analysis, and customized scenarios with higher output quality requirements.
Best for
Specialized high-value scenarios, complex content generation, and quality-sensitive teams
Notes
- Official pricing is per 1K tokens; this page converts to per 1M tokens (×1000). 1500 tokens bills as 1.5K-token units.
- Baichuan2-53B is time-tiered: ¥10/M combined 00:00–8:00, ¥20/M combined 8:00–24:00.
- Knowledge Base API bills Baichuan-Text-Embedding and file storage separately (Embedding ¥0.5/M tokens, storage ¥1.5/GB/day, 5GB cap per user).
Supported coding tools
Open APIWeb SearchAssistants APIKnowledge Base
Pricing and model data sourced from official vendor websites
FAQ
General7 条