Documentation Menu

Birk Models

An AI model family trained by BriqMind on open-source foundations and scaled for enterprise needs. Text, image, and voice capabilities under one roof.

01Birk-Fast

birk-fast-v1
vLLMText + Vision

Near-zero latency for classification, summarization, translation, and instant response systems. It can perform native web search and acts as the brain of the Blink and Blip pipelines.

Latency
~120ms
Context
32K
Num Pred
2K
Serve
vLLM
Fine-tune
Optional
Modality
TextVision Analysis
Native Tools
Web Search (Native)

Runs in a shared pipeline with Blink (image generation) and Blip (voice). It interprets the user prompt and routes it to downstream models.

02Birk-Light

birk-agent-light-v1
SGLangMultimodal

The practical balance between speed and capability. Its JSON output is highly reliable, it uses a limited tool set cleanly, and it is equipped with skills.md reading support plus three primary specialist roles.

Latency
Very Low
Context
64K
Num Pred
8K
Serve
SGLang
Fine-tune
Recommended
Modality
TextVision Analysis
Capabilities
skills.md reading
JSON output - exceptional reliability
3 Primary Specialist Roles (Single-Pass)
Data Analyst

Data querying, visualization, and reporting

Shopping Specialist

Product comparison and recommendation pipeline

Presentation Specialist

End-to-end PowerPoint deck generation

03Birk-Heavy

birk-agent-heavy-v1
SGLangGPU RequiredIndustry Grade

The multimodal flagship. JSON output and error handling are very close to Gemini, Claude, and ChatGPT levels. With a 256K context window, it processes large codebases, reports, and multimedia files end to end. Higher cost, higher power.

Latency
High
Context
256K
Num Pred
8K-16K
Serve
SGLang
Fine-tune
Recommended
Modality (via Pipeline)
Text
Vision Analysis
Voice Create
Voice Analysis
Vision Create
Planning Creator
7 Primary Specialist Roles (Parallel or Sequential)
Data Analyst
Shopping Specialist
Presentation Specialist
Web Search Specialist
Coding Specialist
Math Specialist
Memory-Context Specialist

05Blip

blip-v1
Voice CreateAgenticFast Pipeline

A model that combines voice generation and agentic capabilities. Time to first audio is 0.10s. Chunk-by-chunk streaming provides real-time voice output. Web search and file generation are still actively improving.

First Audio
~0.10s
RLHF
0.10
Streaming
Chunk
Serve
Fast Pipeline
Pipeline Flow
1User writes a prompt
2Birk-Fast produces a text response
3Blip voices it chunk by chunk
In Development
WIPAgentic web search
WIPFile generation as an agent
WIPImproving RLHF score

06Model Comparison

Birk-FastBirk-LightBirk-HeavyBlinkBlip
Serve RuntimevLLMSGLangSGLang
Context Window32K64K256K
Num Pred2K8K8K-16K
Latency~120msVery LowHighVery Fast~0.10s (first)
ModalityText + VisionText + VisionText+Vision+Voice+...Image CreateVoice Create
Tool-callingWeb Search (Native)Limited (reliable)Heavy (industry grade)Web + File (WIP)
Specialist Role3 roles7 roles (parallel)
Pipeline ConnectionBlink + BlipAll modalitiesFastFast
GPU RequirementLowMediumHigh (required)MediumLow

07Benchmark Results

Birk models are evaluated on standard public benchmarks for reasoning, code generation, mathematics, Turkish language understanding, and long-context performance. The results below are produced in BriqMind's internal evaluation environment under zero-shot and equal-prompt conditions; comparison numbers are taken from the relevant models' published technical reports.

Methodology Note
  • All tests were run on the same hardware profile (8x NVIDIA H100 80GB) and inference stack (vLLM 0.6.x / SGLang).
  • Turkish-focused tests used Bogazici University TR-MMLU and TruthfulQA-TR datasets.
  • Scores are reported as the median of three independent runs; measurements with standard deviation above +/-1.2% are marked with a footnote.
  • All test sets and evaluation prompts are available on request for verification.

Academic and General Capability Benchmarks

BenchmarkMeasureBirk-HeavyBirk-LightBirk-FastGPT-4oClaude 3.5 SonnetGemini 1.5 Pro
MMLUGeneral reasoning, 5-shot
MMLU-ProMore difficult MMLU
GPQA DiamondGraduate-level science
HellaSwagCommonsense inference
ARC-ChallengeScience questions
BBHBIG-Bench Hard

Code and Mathematics Capabilities

BenchmarkMeasureBirk-HeavyBirk-LightGPT-4oClaude 3.5 Sonnet
HumanEvalPython code generation, pass@1
MBPPMulti-language code generation
LiveCodeBenchCompetitive programming problems
GSM8KGrade-school mathematics
MATHHigh-school mathematics
AIME 2024Olympiad mathematics

Turkish Language Capabilities

Turkish natural language tests are one of the most critical categories for local buyers and public-sector stakeholders. Birk models are specifically fine-tuned for Turkish understanding, generation, and cultural context.

BenchmarkMeasureBirk-HeavyBirk-LightBirk-FastGPT-4oClaude 3.5 Sonnet
TR-MMLUTurkish general reasoning
TruthfulQA-TRTurkish truthfulness evaluation
Belebele (TR)Turkish reading comprehension
XCOPA (TR)Turkish commonsense inference
Turkish HellaSwagTurkish commonsense completion
Turkish Spelling & GrammarBriqMind internal evaluation set
Enterprise Turkish Q&AFinance / legal / manufacturing domain tests

Long-Context and Agent Performance

BenchmarkMeasureBirk-HeavyBirk-LightComparison
Needle-in-a-Haystack (128K)Information retrieval in long context
RULER (200K)Multi-step long-context reasoning
BFCLBerkeley Function Calling Leaderboard
ToolBenchMulti-step tool use
AgentBenchAutonomous agent task completion
JSON Schema ComplianceStructured output accuracy

Inference Performance

MetricBirk-FastBirk-LightBirk-Heavy
Time to first token (TTFT)
Output token speed (tokens/sec)
Concurrent user capacity (reference hardware)
Context prefill throughput (K-token/sec)
P50 end-to-end response time (1K context)
P99 end-to-end response time (1K context)
Reproducibility: For every benchmark, the seed, prompt template, output parser, and evaluator version are fixed in the BriqMind technical report. Independent verification requests can be sent to research@briqmind.com

08API Usage

cURL
curl https://api.briqmind.com/v1/chat/completions \ -H "Authorization: Bearer $BRIQ_API_KEY" \ -d '{ "model": "birk-agent-heavy-v1", "messages": [{ "role": "user", "content": "..." }] }'

You can choose a model by changing the model parameter. If orchestration is required, manage it in the application or agent workflow layer; the model parameter should receive the real model name.