Trusted by industry leaders.Used by the CEOS.
Changelog from Blaze Research Labs
January 2025
January 2025
Initiated the MVP of Blaze — a deep research engine designed to intelligently fetch, analyze, and surface verified knowledge from the web.
Started building the foundational backend infrastructure with scraping, vector storage, and citation tracing capabilities.
February 2025
February 2025
Released internal alpha of Blaze — capable of querying multiple sources with basic context preservation.
Integrated semantic search and source-quality scoring to prioritize credible content.
March 2025
March 2025
Deployed Blaze Beta with:
- Boosted source crawling accuracy using domain-specific heuristics
- Added initial version of Blaze Reasoner (v1) — a model for summarizing and fact-checking results
- Implemented session-based memory to retain context across research threads
April 2025
April 2025
As of April 6, 2025 — Blaze now powers live deep research queries with:
- Realtime source ranking and live web data integration
- Advanced Blaze Reasoner v2 with argument synthesis and contradiction detection
- Full-text traceability for every citation in the final answer
- Contextual follow-up system (auto-suggested sub-questions)
- Currently scoring 32% on humanity's last exam, with projections reaching 41.2% soon
Coming Soon 🚧
Coming Soon 🚧
Upcoming features and improvements for Blaze include:
- Enhanced multi-language support for global research capabilities
- Blaze Reasoner v3 with improved reasoning and predictive analytics
- Integration with external APIs for real-time data enrichment
Accessing worldwide information sources to deliver comprehensive and accurate research results.
AI-powered research providing comprehensive analysis and synthesized insights on complex topics.
Tracking the evolution of AI capabilities over time. Blaze Deep Research leads with a score of 33% in the latest benchmark.
Transforming raw business data into actionable insights with real-time performance tracking.
Humanity's Last Exam
Final Benchmark Challenges For AI Models
Recent Models
Blaze Deep Research
33%
OpenAI Deep Research
28%
O3 Mini High
11%
Blaze Deep Research Engine - Benchmark Performance
Performance progression on key AI benchmarks (HumanEval & GPQA).
HumanEval Benchmark
Code generation capability assessment.
GPQA Benchmark
Graduate-level science question answering.
Ready to Transform Your Research?
Join thousands of researchers who've already revolutionized their discovery process with Blaze.