DBRX

DBRX was a credible 2024 open MoE release, with Databricks reporting 70.1% HumanEval pass@1 and an official repository supporting local developer use. However, the available independent evidence is dated: Artificial Analysis assigned an Intelligence Index of 3 in March 2024, and Arena-Hard-Auto v0.1 reports 23.9. It is text-oriented in the supplied evidence, with no supported vision, audio, tool-use, or structured-I/O capability claim. Relative to published anchors, DBRX is materially below Claude Opus 4.8, GPT-5.5, and GPT-5.6 Sol on current capability, reasoning, and agentic coding evidence. It also trails Kimi K2.5 and DeepSeek-V3.2, which have substantially stronger contemporary coding or general-capability profiles. GLM-4.6 is the closest lower-tier comparison, but DBRX's historical HumanEval result supports a modest coding baseline while not establishing modern repository-agent performance. Against Gemma 3, DBRX has stronger historical code evidence but is clearly worse on multimodal coverage and has less current operational evidence. Open availability can reduce model-license acquisition cost, and the historical Artificial Analysis listing showed zero hosted-token price, but neither fact establishes present hosted pricing or total deployment cost. Artificial Analysis reports no current provider speed or latency data. DeepSWE and current LiveCodeBench contain no DBRX entry; SWE-bench evidence was not found. A secondary CloudPrice listing reports LiveCodeBench 0.1 and rank #301, but this conflicts with the official leaderboard's absence and should not be treated as a robust benchmark result.