Articles by tag: llm-benchmark

Benchmark of LLM Servers for Qwen 3.5 on Mac — 2026 Results

Comparison of 8 local LLM servers for Qwen 3.5 on Mac. Find out which framework gives 2.17x acceleration with parallel requests. Recommendations for developers.

LLM Benchmark: GPT-5.4 vs Claude Opus 4.6

LLM Persuasion Benchmark: GPT-5.4 is more persuasive than all in 6300 debates. Persuasiveness and resilience ratings. Analysis for developers — study the data and apply in projects.

EVGeoQA: LLM test in geospatial search

Explore EVGeoQA — benchmark for dynamic EV charging search with dual objectives. LLM evaluation, GeoRover, weaknesses at long distances. Download the dataset for tests.

Kimi K2.5 competes with GPT-5.4 without VPN

Testing 54 LLM: Kimi K2.5 and MiniMax lead among those available from Russia. Comparison on managers' tasks, benchmarks, examples. Choose a model for analytics and communication.

EduBench-RU: test of 22 LLMs for schools in Russia

EduBench-RU benchmark results: Gemini leads, Chuvash not mastered by any model. Testing FGOS, OGE, KTP. Download prompts and test your LLMs.

NSFW filter on LLM: test Phi-3, Qwen, Gemma

Comparison of LLM for filtering NSFW prompts before image generation. Accuracy up to 93%, latency from 85 ms. Model benchmarks and Grammar. Implement in your project.