Models

Pick the perfect model

Frontier models tuned for reliability, speed, and cost. Switch anytime with one line.

Bcas Lite

100 credits
Everyday
Speed
Instant
Quality
Very Good
Context
128K
Best for
High-volume chatClassification
Buy credits

Bcas Pro

200 credits
Most popular
Speed
Fast
Quality
Excellent
Context
200K
Best for
General chatSupportSales
Buy credits

Bcas Vision

500 credits
Multimodal
Speed
Fast
Quality
Excellent
Context
128K
Best for
ImagesDocumentsOCR
Buy credits

Bcas Ultra

800 credits
Flagship
Speed
Balanced
Quality
State-of-the-art
Context
1M
Best for
ReasoningAgentsAnalysis
Buy credits

Bcas Reason

1,200 credits
Deep think
Speed
Deliberate
Quality
Expert
Context
500K
Best for
Complex logicCodingMath
Buy credits

Bcas Voice

2,000 credits
Real-time
Speed
Real-time
Quality
Natural
Context
—
Best for
Phone callsVoice agents
Buy credits
Benchmarks

Measured, not marketed

Public evaluations, run monthly and posted with methodology.

UltraProLite
MMLU (knowledge)higher is better
92
87
78
HumanEval (code)higher is better
89
82
66
GSM8K (math)higher is better
95
90
74
Latency P50 (ms)lower is better
620
340
180
Capabilities

Everything the platform can do

A shared runtime across every Bcasfintech model.

Streaming chat

Server-sent events with sub-second first token.

Tool calling

Function calling, retrieval, and multi-step agents.

Long context

Up to 1M tokens for whole-repo or corpus-scale prompts.

Vision & docs

Read PDFs, screenshots, charts, and handwriting.

Voice I/O

Turn-taking, barge-in, and natural prosody.

Guardrails

PII redaction, safety filters, and content policies.

Recommendations

Which model for your job?

A quick map from real workloads to the right model.

Use case
Recommended
Why
Customer support
Bcas Pro
Fast, warm tone, handles 90% of tier-1 tickets.
Sales outreach
Bcas Pro
Balanced speed and creativity for personalized outreach.
Legal & research
Bcas Reason
Expert step-by-step reasoning with citations.
Coding copilot
Bcas Reason
Top-tier HumanEval scores and long-context repo review.
Document intake
Bcas Vision
OCR, table extraction, structured JSON output.
Phone concierge
Bcas Voice
Real-time, low-latency conversational voice.
99.99%
Uptime SLA (Ultra)
<200ms
First-token latency (Lite)
6
Regions worldwide
SOC 2
Type II certified
Data isolation per workspace · encrypted in transit and at rest · optional zero-retention mode.
FAQ

Common questions

How is credit usage calculated?

One credit is $0.10. Each model has a per-1K-token rate — Lite is 100 credits per 1K, Ultra is 800. Voice is billed per minute of audio.

Can I switch models mid-conversation?

Yes. Pass a different model in the next request. Conversation history and tool state are preserved.

Do you fine-tune?

Ultra and Pro support LoRA fine-tuning on your own data. Contact us to enable it.

Is my data used for training?

No. Your prompts and outputs are never used to train foundation models unless you explicitly opt in.