Quasar 438B Release: Europe's Highest-Scoring Reasoning Model

Quasar 438B is the highest-scoring European model on the Artificial Analysis Intelligence Index

Multiverse Computing has released Quasar 438B, a reasoning model designed for enterprise-scale agents and coding. The model scores 43 on the Artificial Analysis Intelligence Index v4.1.1, the highest result for any European model in that benchmark. It is designed to handle multi-step tasks requiring planning, tool use, code execution, and large context windows, and is available via the CompactifAI API.

Performance and Intelligence Benchmarks

Quasar 438B's score of 43 on the Artificial Analysis Intelligence Index (a composite of nine evaluations including GPQA Diamond and SciCode) places it ahead of several other models, including:

Inkling: 42
NVIDIA Nemotron 3 Ultra: 38
Mistral Medium 3.5: 30

While it leads these models, it remains behind frontier models such as Claude Opus 5, which leads the index with a score of 63.

Latency and Response Speed

Quasar 438B is optimized for low latency in the 400B+ parameter class. It returns 500 tokens (including thinking time) in 15.3 seconds. In comparison to other high-scoring models:

Mistral Medium 3.5: 18.8 seconds (Score: 30)
NVIDIA Nemotron 3 Ultra: 25.7 seconds (Score: 36)
Inkling: 48.3 seconds (Score: 42)

Only three models in the comparison are faster: Nemotron 3.5 Lightning (9.4s), Gemini 3.5 Flash-Lite (10.8s), and Gemini 3.7 Flash (11.5s). Of these, only Gemini 3.7 Flash both outperforms Quasar on the intelligence index and provides faster response times.

Long-Context and Agentic Capabilities

Quasar 438B demonstrates strong performance in long-context reasoning and terminal-based agentic work:

Long-Context Reasoning (AA-LCR): Quasar scores 75.0, placing it on par with Grok 4.6 (high) and within one point of Claude Opus 5 (75.7) and Qwen3.8 2.4T A95B (75.3).
Agentic Coding (Terminal-Bench v2.1): Quasar scores 69.3, leading Mistral Medium 3.5 by 18.7 points and Nemotron 3 Ultra by 15.4 points, though it trails Claude Opus 5 (89.1).

Community Critique and Technical Skepticism

Following the announcement, technical discussions on Hacker News raised several concerns regarding the model's origins, transparency, and the company's claims.

Model Provenance and Transparency

Critics pointed to a lack of transparency regarding whether the model was pretrained from scratch or is a compressed/fine-tuned version of an existing model. Some users suggested it may be a modified version of GLM 5.2 or MiniMax M3 based on parameter counts and changelog entries.

"I think this is GLM 5.2 with parameters removed. Its advertised in their changelog as 'capabilities are identical to GLM 5.2,' it has the same two effort settings 'high' and 'max'"

Skepticism of "Quantum AI" Claims

Multiverse Computing's use of "quantum physics" and "tensor networks" for model compression via their CompactifAI technology has been met with skepticism. Users questioned the utility of quantum algorithms in current LLM development and characterized some of the company's marketing as "technobabble."

Closed Weights and Benchmarking

Because Quasar 438B is API-only and the weights are not public, some developers expressed distrust in the provided benchmarks, noting that "frontier" models often include hidden system prompts or wrappers that inflate scores compared to "naked" open-weight models.

"When the weights are closed I don't believe any benchmark... 'Frontier' models get tested as a model + whatever secret sauce they choose to put in front."

Deployment and Availability

Quasar 438B supports English and Spanish and is accessible through the CompactifAI API, allowing enterprise teams to integrate the model into software development agents, technical copilots, and research systems without managing their own infrastructure.