Commercial placements are independent of Official Core; a commercial relationship cannot buy test conclusions.Read the disclosure rules →

PARTNER PROFILE

B.AI

Third-party API gateway (OpenAI-compatible)

Direct connection to the official site API · 44 models covered · some models free to use

SERVICE SNAPSHOT

Billing modelUsage-based billing

CurrencyUSD

Last verified2026-08-27

MODELS

44 public models (refer to the official page)
DeepSeek V4-Flash
Xiaomi MiMo V2.5
Tencent Hy3

API COMPATIBILITY

OpenAI Compatible (configuration test protocol)

REGION

Unknown

PAYMENT

Top-up (specific methods: refer to the official page)

SUITABLE FOR

Developers who need an OpenAI-compatible interface
Developers who want to try low-cost or free models

LIMITATIONS

The scope and availability of free models may change
Specific rate limits, regions and payment conditions are not fully verified

PUBLIC OFFER

Some models free to use

Partially Verified

The official price table shows input/output/cache at CNY 0 for DeepSeek V4-Flash, Xiaomi MiMo V2.5 and Tencent Hy3; all 44 models supported (verified 2026-08-27)

Amount: 0 (some models) CNY

Condition: Applies only to models marked at CNY 0 in the public price table, Specific availability: refer to the Provider's official page

Expiry: Not Published

Benefits can change; the provider's official page prevails.

PRICING NOTE

Some models at CNY 0 (DeepSeek V4-Flash / Xiaomi MiMo V2.5 / Tencent Hy3)

DeepSeek V4-Pro approximately $1.32 (input) / $3.96 (output) per 1M tokens; top-up tiers at 10%-40% off; refer to the official page

DATA SOURCES

Provider website ↗Verified Unknown

OFFICIAL LINKS

Official documentation: Not Published
Official console ↗

INDEPENDENT TEST EVIDENCE

The test status on this page describes only the listed test conditions and samples. It is not official certification, a performance guarantee or a Provider ranking. Official Core's independent methodology and full testing entry points are published separately.

Read the independent methodology ↗Open authorized testing ↗

BENCHMARK FACTS

APISpotlight Tested

Machine tested

All rows use an OpenAI-compatible chat completion request; latency is the P50 of successful samples and the success rate is computed over this run's samples.

Model	Status	P50 latency	Success rate	Tested at	Result
deepseek-v4-flashdeepseek-v4-flash	Third-party	2305 ms	100%	Aug 28, 2026, 12:34 AM	Passed

APISpotlight Tested: our own machine-test factsPartner Reported: information published or supplied by the PartnerNot Tested: no verifiable test result from this site

SAME CANONICAL MODEL

Official vs Third-party

No ranking

Side-by-side measured facts for the same canonical model across first-party and third-party providers. No composite score is calculated and no ordering is implied.

Provider	Type	Latency	Success rate	Tested at	Status
DeepSeekDeepSeek V4 Flash · deepseek-v4-flash	Official	410 ms	100%	Sep 29, 2026, 4:33 PM	Passed
B.AIdeepseek-v4-flash	Third-party	2305 ms	100%	Aug 28, 2026, 12:34 AM	Passed