GLM-5.2

SiliconFlow published a developer-focused comparison titled "Best Cheap API for SillyTavern: DeepSeek-V4-Flash, GLM-5.2, and Long-Context Cost" on August 13, 2026. The article frames GLM-5.2 as one of two paths for long SillyTavern sessions: a low-cost option (DeepSeek-V4-Flash) versus a model designed for more demandi The piece also explains how context growth affects token usage in SillyTavern, listing elements such as character card, scenario, example dialogue, chat history, World Info/lorebook, user persona, author's note, and model output. It notes that a 1049K context model charges only for tokens actually processed and generat

SiliconFlow's launch post "GLM-5.2 Now on SiliconFlow: 1M Context, Long-Horizon Engineering, Near-Opus Coding" (June 24, 2026) announces GLM-5.2 as Z.ai's latest open-source flagship going live on SiliconFlow. The post cites frontier-level coding and agentic performance, claiming GLM-5.2 matches or beats GPT-5.5 on ben At launch, SiliconFlow listed GLM-5.2 with a 1,049K-token context window, $1.40/M input, $4.40/M output, and $0.26/M cache read pricing, served via FP8 precision with function calling, context caching, thinking, and dual reasoning effort (High and Max) capabilities. The post highlights OpenAI- and Anthropic-compatible

A third-party guide updated September 9, 2026, separates independently verified GLM-5.2 benchmark results from Zhipu's self-reported numbers across categories. GLM-5.2's clearest independently verified result is its #1 finish on Design Arena's Code Categories leaderboard, a blind human-preference test where it ranks ro The guide notes that specific point scores on SWE-bench Pro (62.1), Terminal-Bench 2.1, and FrontierSWE come from Zhipu's own technical report and model card — directionally corroborated by independent evaluators but not reproduced exactly by a third party using Zhipu's precise methodology. The headline framing of 'nea

OpenRouter's GLM 5.2 directory page describes the model as a large-scale reasoning model from Z.ai supporting text input/output with a 1M-token context window, suited to long-horizon agent workflows, project-level software engineering, and complex multi-step automation. It notes that reasoning efforts "high" and "xhigh The page catalogs 20+ hosting providers with per-million-token input, output, and cache-read pricing, plus latency, throughput, and uptime metrics, and lets users filter by quantization and routing mode (Balanced, Nitro, Exacto). SiliconFlow appears with a 15% discount at $1.19 input / $3.74 output / $0.221 cache read

pricepertoken.com's GLM 5.2 API Pricing 2026 page (last updated September 1, 2026) reports that GLM 5.2 was released on June 16, 2026, with pricing starting at $0.402 per million input tokens and $1.26 per million output tokens, plus a $0.060 per 1M cached token rate. The model is listed with a 1,048,576-token context The page highlights a 71.3% drop in GLM 5.2's cheapest input price over the past 90 days, falling from $1.40 to $0.402 per million tokens, signaling aggressive price competition across providers. It lists 20 providers for the model, with FlexAI cited as the cheapest at $0.402 per million input tokens and input prices r

Z.ai (formerly Zhipu AI) released GLM-5.2 on June 16, 2026, as a 753-billion-parameter open-weights large language model engineered for long-horizon autonomous coding and engineering tasks, according to VentureBeat reporting. The model ships under an MIT open-source license on Hugging Face, is also available via the Z. Architecturally, GLM-5.2 introduces an optimization called IndexShare, which reuses a single indexer across every four sparse attention layers, reducing per-token compute FLOPs by approximately 2.9x at the maximum 1-million-token context length. It also features an upgraded Multi-Token Prediction layer for speculative

NIST's Center for AI Standards and Innovation published an assessment of Z.ai's GLM-5.2 on July 8, 2026, reporting it was probably the most capable open-weight model at its June 16, 2026 release. CAISI found GLM-5.2's overall capabilities are similar to GPT-5.2 (December 2025) and its cyber capabilities are similar to Anthropic's Opus 4.6 (February 2026). The evaluation also flagged mixed safeguards performance: GLM-5.2 allowed assistance with agentic cyber exploit development, blocked fewer sensitive biological questions than reference U.S. models, but appeared potentially more robust against agent hijacking and jailbreaking than other evaluated PRC open-weight models. The report notes that prompt-based safeguards can still be circumvented when the open-weight model is self-hosted, a key caveat for developers.

An aggregator page presents all 17 published GLM-5.2 benchmark results grouped into reasoning, coding, and agentic categories, explicitly disclaiming that the figures are republished as published by the model authors and that the site does not run the evaluations. The long-horizon trio — FrontierSWE (74.4%, up to 20-ho Generation-over-generation deltas versus GLM-5.1 show uneven gains: +3.7 on SWE-bench Pro (62.1 vs 58.4), +17.5 on Terminal-Bench 2.1 (81.0 vs 63.5), +28.2 (2.6×) on DeepSWE (46.2 vs 18.0), +12.8 on ProgramBench (63.7 vs 50.9), +5.0 on MCP-Atlas, +7.5 on Tool-Decathlon (48.2 vs 40.7), and +9.5 on HLE (40.5 vs 31.0). Th

Artificial Analysis's "GLM-5.2 (Non-reasoning): API Provider Performance Benchmarking & Price Analysis" evaluates the model across 8 API providers. Output speed rankings put Baseten (FAST) at 215.9 t/s first, Modular at 165.5 t/s second, Nebius (FP4) at 118.9 t/s third, Baseten at 105.3 t/s fourth, and SiliconFlow (FP8 For pricing, Bitdeer AI leads at $0.37 blended per 1M tokens, followed by Wafer at $0.79, Baseten at $0.82, SiliconFlow (FP8) at $0.85, and Modular at $0.90, with up to a 6.8x spread across providers. The analysis also breaks down separate cache hit, input, and output prices per provider and notes that cache write and