AIPosted 10/12/202610/12/2026 at 9:12 pm@TechDesk

Starts 06/16/2026CONFIRMED84% CONFIDENCE

Z.ai's Hugging Face^[6] repository, API and developer release notes all date GLM-5.2 to 16 June 2026 and Epoch^[5] AI lists "June 16, 2026". Coding Plan subscribers had it from 13 June (a Saturday); AI Weekly^[2] says that launch promised the open weights and API for 16 June, so 13 June was a staged first access, not the full release.

Start: Jun 16, 202692% CONFIDENCEfrom epoch.ai

GLM-5.2 Released

Z.ai (Zhipu AI) releases GLM-5.2, an MIT-licensed 753B-parameter flagship built for long-horizon coding with a 1M-token context window, scoring 62.1 on SWE-bench Pro and 81.0 on Terminal Bench 2.1.

Z.ai, the international brand of Zhipu AI, opened GLM-5.2 to every GLM Coding Plan tier on 13 June 2026 and on 16 June added its API, chat access and MIT-licensed weights on Hugging Face^[2][4][6]
It is pitched as the flagship "for the era of long-horizon tasks": a 1M-token context window that Z.ai says is genuinely usable, five times the 200K of GLM-5.1, with fewer context drifts and forgotten goals on project-scale work^[1][2]
Z.ai's table has it at 62.1 on SWE-bench Pro (GLM-5.1 58.4, Claude Opus 4.8 69.2), 81.0 on Terminal Bench 2.1 (Opus 4.8 85.0), 74.4 on FrontierSWE against Opus 4.8's 75.1 and 13.0 on SWE-Marathon against 26.0^[6]
Independent checks were friendlier: Artificial Analysis gave it 51 on its Intelligence Index v4.1, ahead of MiniMax-M3 and DeepSeek V4 Pro at 44, and it ranked second on Code Arena with 1,595 points^[3]
Price was part of the pitch: about $1.40 per million input and $4.40 per million output tokens against $5 and $30 for GPT-5.5^[3]
Trending Topics notes the open-source release was announced the same day the Trump administration ordered Anthropic's Fable 5 and Mythos 5 blocked for foreign nationals, and that Z.ai's Hong Kong-listed shares jumped as much as 48% at the start of that week^[3]

Notable features

Text input and output, 1M-token context and 128K-token maximum output, with two selectable reasoning efforts, High and Max^[1][4]
Mixture-of-experts model with about 753 billion total parameters (press reports quote 744B total and about 40B active) under the MIT licence^[3][6]
IndexShare reuses one indexer across every four sparse-attention layers, cutting per-token compute 2.9x at a 1M context, and an improved multi-token-prediction layer lifts speculative-decoding acceptance by up to 20%^[6]
Runs locally on SGLang, vLLM, Transformers, KTransformers and Unsloth, with Ascend NPU support through vLLM-Ascend, xLLM and SGLang^[6]
Z.ai's docs steer users to project-level codebase takeover, long refactors in /goal mode and multi-platform builds as the tasks it was tuned for^[1]

Benchmarks

Z.ai's results table from the model card, as published^[6]; * marks results on the full test set.

Benchmark	GLM-5.2	GLM-5.1	Qwen3.7-Max	MiniMax M3	DeepSeek-V4-Pro	Claude Opus 4.8	GPT-5.5	Gemini 3.1 Pro
Reasoning
HLE	40.5	31	41.4	37	37.7	49.8*	41.4*	45
HLE (w/ Tools)	54.7	52.3	53.5	-	48.2	57.9*	52.2*	51.4*
CritPt	20.9	4.6	13.4	3.7	12.9	20.9	27.1	17.7
AIME 2026	99.2	95.3	97	-	94.6	95.7	98.3	98.2
HMMT Nov. 2025	94.4	94	95	84.4	94.4	96.5	96.5	94.8
HMMT Feb. 2026	92.5	82.6	97.1	84.4	95.2	96.7	96.7	87.3
IMOAnswerBench	91.0	83.8	90	-	89.8	83.5	-	81
GPQA-Diamond	91.2	86.2	90	93	90.1	93.6	93.6	94.3
Coding
SWE-bench Pro	62.1	58.4	60.6	59	55.4	69.2	58.6	54.2
NL2Repo	48.9	42.7	47.2	42.1	35.5	69.7	50.7	33.4
DeepSWE	46.2	18	18	20	8	58	70	10
ProgramBench	63.7	50.9	-	-	47.8	71.9	70.8	39.5
Terminal Bench 2.1 (Terminus-2)	81.0	63.5	75	65	64	85	84	74
Terminal Bench 2.1 (Best Reported Harness)	82.7	69	-	-	-	78.9	83.4	70.7
FrontierSWE (Dominance)	74.4	30.5	-	-	29.0	75.1	72.6	39.6
PostTrainBench	34.3	20.1	-	-	-	37.2	28.4	21.6
SWE-Marathon	13.0	1.0	-	-	-	26.0	12.0	4.0
Agentic
MCP-Atlas (Public Set)	76.8	71.8	76.4	74.2	73.6	77.8	75.3	69.2
Tool-Decathlon	48.2	40.7	-	-	52.8	59.9	55.6	48.8

@TechDesk

References 684% CONFIDENCE

The first entry is always the pin's source. Overall confidence is a weighted average of how firmly each reference supports the start and end times used above; a reference counts half as much for every 180 days older than the newest.

[1]90%

docs.z.ai/guides/llm/glm-5.2

docs.z.ai· Posted Oct 12, 2026· Starts Jun 16, 2026· 24% of score

[2]72%

Z.ai's GLM-5.2 Brings 1M-Token Context to Open-Weight Coding - AI Weekly

aiweekly.co· Published Jun 29, 2026· 16% of score

Records the 13 June Coding Plan launch with MIT-licensed weights and standalone API access promised for 16 June.^[1]

[3]80%

GLM-5.2: China's Zhipu AI Beats Even Google's Top Models With Its New Open LLM - Trending Topics

trendingtopics.eu· Published Jun 18, 2026· 15% of score

Independent context: Artificial Analysis Intelligence Index 51 (MiniMax-M3 44, DeepSeek V4 Pro 44, Kimi K2.6 43), second on Code Arena with 1,595 points, about $1.40 and $4.40 per million tokens, the 13 June Coding Plan rollout and the stock reaction.^[1]

[4]75%

GLM-5.2: Z.ai's Open-Weights Coder Beats GPT-5.5 at 1/6 the Cost - NYU Shanghai RITS

rits.shanghai.nyu.edu· Published Jun 18, 2026· 15% of score

Explains the High and Max effort levels, the 2.9x per-token compute saving from IndexShare, and that the API and chatbot followed alongside the open release.^[1]

[5]92%

GLM-5.2 - Epoch AI

epoch.ai· Published Jun 16, 2026· Starts Jun 16, 2026 ✓· 15% of score

Epoch AI's model database lists GLM-5.2 as an open-weights model from Z.ai released 16 June 2026, ranking 20th of 223 tracked models at release.

Suggest a correction

Something missing or wrong? Say it in your own words: a link that backs this pin up, a different start or end date and why, or a fact it lacks or gets wrong. The AI checks it against this pin's sources, searches for better ones, and adds any page that backs you up. The pin's own sources still count most. A picture that shows something else, or shows it badly, is looked at too, and moved down or replaced.

GLM-5.2 Released

Z.ai (Zhipu AI) releases GLM-5.2, an MIT-licensed 753B-parameter flagship built for long-horizon coding with a 1M-token context window, scoring 62.1 on SWE-bench Pro and 81.0 on Terminal Bench 2.1.

Z.ai, the international brand of Zhipu AI, opened GLM-5.2 to every GLM Coding Plan tier on 13 June 2026 and on 16 June added its API, chat access and MIT-licensed weights on Hugging Face^[2][4][6]
It is pitched as the flagship "for the era of long-horizon tasks": a 1M-token context window that Z.ai says is genuinely usable, five times the 200K of GLM-5.1, with fewer context drifts and forgotten goals on project-scale work^[1][2]
Z.ai's table has it at 62.1 on SWE-bench Pro (GLM-5.1 58.4, Claude Opus 4.8 69.2), 81.0 on Terminal Bench 2.1 (Opus 4.8 85.0), 74.4 on FrontierSWE against Opus 4.8's 75.1 and 13.0 on SWE-Marathon against 26.0^[6]
Independent checks were friendlier: Artificial Analysis gave it 51 on its Intelligence Index v4.1, ahead of MiniMax-M3 and DeepSeek V4 Pro at 44, and it ranked second on Code Arena with 1,595 points^[3]
Price was part of the pitch: about $1.40 per million input and $4.40 per million output tokens against $5 and $30 for GPT-5.5^[3]
Trending Topics notes the open-source release was announced the same day the Trump administration ordered Anthropic's Fable 5 and Mythos 5 blocked for foreign nationals, and that Z.ai's Hong Kong-listed shares jumped as much as 48% at the start of that week^[3]

Notable features

Text input and output, 1M-token context and 128K-token maximum output, with two selectable reasoning efforts, High and Max^[1][4]
Mixture-of-experts model with about 753 billion total parameters (press reports quote 744B total and about 40B active) under the MIT licence^[3][6]
IndexShare reuses one indexer across every four sparse-attention layers, cutting per-token compute 2.9x at a 1M context, and an improved multi-token-prediction layer lifts speculative-decoding acceptance by up to 20%^[6]
Runs locally on SGLang, vLLM, Transformers, KTransformers and Unsloth, with Ascend NPU support through vLLM-Ascend, xLLM and SGLang^[6]
Z.ai's docs steer users to project-level codebase takeover, long refactors in /goal mode and multi-platform builds as the tasks it was tuned for^[1]

Benchmarks

Z.ai's results table from the model card, as published^[6]; * marks results on the full test set.

Benchmark	GLM-5.2	GLM-5.1	Qwen3.7-Max	MiniMax M3	DeepSeek-V4-Pro	Claude Opus 4.8	GPT-5.5	Gemini 3.1 Pro
Reasoning
HLE	40.5	31	41.4	37	37.7	49.8*	41.4*	45
HLE (w/ Tools)	54.7	52.3	53.5	-	48.2	57.9*	52.2*	51.4*
CritPt	20.9	4.6	13.4	3.7	12.9	20.9	27.1	17.7
AIME 2026	99.2	95.3	97	-	94.6	95.7	98.3	98.2
HMMT Nov. 2025	94.4	94	95	84.4	94.4	96.5	96.5	94.8
HMMT Feb. 2026	92.5	82.6	97.1	84.4	95.2	96.7	96.7	87.3
IMOAnswerBench	91.0	83.8	90	-	89.8	83.5	-	81
GPQA-Diamond	91.2	86.2	90	93	90.1	93.6	93.6	94.3
Coding
SWE-bench Pro	62.1	58.4	60.6	59	55.4	69.2	58.6	54.2
NL2Repo	48.9	42.7	47.2	42.1	35.5	69.7	50.7	33.4
DeepSWE	46.2	18	18	20	8	58	70	10
ProgramBench	63.7	50.9	-	-	47.8	71.9	70.8	39.5
Terminal Bench 2.1 (Terminus-2)	81.0	63.5	75	65	64	85	84	74
Terminal Bench 2.1 (Best Reported Harness)	82.7	69	-	-	-	78.9	83.4	70.7
FrontierSWE (Dominance)	74.4	30.5	-	-	29.0	75.1	72.6	39.6
PostTrainBench	34.3	20.1	-	-	-	37.2	28.4	21.6
SWE-Marathon	13.0	1.0	-	-	-	26.0	12.0	4.0
Agentic
MCP-Atlas (Public Set)	76.8	71.8	76.4	74.2	73.6	77.8	75.3	69.2
Tool-Decathlon	48.2	40.7	-	-	52.8	59.9	55.6	48.8

References 684% CONFIDENCE

[1]90%

docs.z.ai/guides/llm/glm-5.2

docs.z.ai· Posted Oct 12, 2026· Starts Jun 16, 2026· 24% of score

[2]72%

Z.ai's GLM-5.2 Brings 1M-Token Context to Open-Weight Coding - AI Weekly

aiweekly.co· Published Jun 29, 2026· 16% of score

Records the 13 June Coding Plan launch with MIT-licensed weights and standalone API access promised for 16 June.^[1]

[3]80%

GLM-5.2: China's Zhipu AI Beats Even Google's Top Models With Its New Open LLM - Trending Topics

trendingtopics.eu· Published Jun 18, 2026· 15% of score

[4]75%

GLM-5.2: Z.ai's Open-Weights Coder Beats GPT-5.5 at 1/6 the Cost - NYU Shanghai RITS

rits.shanghai.nyu.edu· Published Jun 18, 2026· 15% of score

Explains the High and Max effort levels, the 2.9x per-token compute saving from IndexShare, and that the API and chatbot followed alongside the open release.^[1]

[5]92%

GLM-5.2 - Epoch AI

epoch.ai· Published Jun 16, 2026· Starts Jun 16, 2026 ✓· 15% of score

Epoch AI's model database lists GLM-5.2 as an open-weights model from Z.ai released 16 June 2026, ranking 20th of 223 tracked models at release.

Suggest a correction

GLM-5.2 Released

Notable features

Benchmarks

Suggest a correction

GLM-5.2 Released

Notable features

Benchmarks

Suggest a correction

Tags

Thread