BlogLLM Benchmark Scores as Published, Side by Side

Vending-Bench: Results as Published

Every result the catalog holds for Vending-Bench: 17 models from 7 developers, read from 3 pages. Each number is as printed by the company named beside it; nothing is ranked or rescored. Models are listed by developer, newest first.

Area Knowledge WorkMaintainer Andon LabsIntroduced 2025Data version 2026103001, as of 2026-10-30

Vending-Bench puts a model in charge of a simulated vending-machine business and scores the money left in the bank. On the version 2 page, the model starts with $500, has one simulated year, and is stopped early if it cannot pay the $2 daily machine fee for more than 10 days in a row. It finds suppliers on the web, orders by email, moves stock between storage and the machine, and sets prices, while sales depend on the day, season, weather, and price. Version 2 also lets suppliers act adversarially, delay deliveries, or disappear, and lets customers demand refunds.

Vending-Bench 2 (final balance in USD)

ModelValuePrinted bySetup as printed
GPT-6 AstraOpenAI, released 2026-09-03$15515Google, Gemini 4 Carbonchecked 2026-10-29-
GPT-5.4OpenAI, released 2026-03-05$6144.18Z.ai, GLM-5.1 model cardchecked 2026-10-15-
GPT-5.2OpenAI, released 2025-12-11$3591.33Z.ai, GLM-5 model cardchecked 2026-10-15Effort: xhigh
Claude Opus 5.5Anthropic, released 2026-09-22$9235Google, Gemini 4 Carbonchecked 2026-10-29-
Claude Fable 5.1Anthropic, released 2026-09-01$5422Google, Gemini 4 Carbonchecked 2026-10-29-
Claude Opus 4.6Anthropic, released 2026-02-05$8017.59Z.ai, GLM-5.1 model cardchecked 2026-10-15-
Claude Opus 4.5Anthropic, released 2025-11-24$4967.06Z.ai, GLM-5 model cardchecked 2026-10-15-
Gemini 4 CarbonGoogle, released 2026-10-28$16484Google, Gemini 4 Carbonchecked 2026-10-29-
Gemini 4 ArgonGoogle, released 2026-09-30$13718Google, Gemini 4 Carbonchecked 2026-10-29-
Gemini 3.1 Pro PreviewGoogle, released 2026-02-19$911.21Z.ai, GLM-5.1 model cardchecked 2026-10-15-
Gemini 3 ProGoogle, released 2025-11-18$5478.16Z.ai, GLM-5 model cardchecked 2026-10-15-
DeepSeek-V3.2DeepSeek, released 2025-12-01$1034Z.ai, GLM-5 model cardchecked 2026-10-15; printed on 1 more page-
Qwen3.6-PlusAlibaba, released 2026-04-02$5114.87Z.ai, GLM-5.1 model cardchecked 2026-10-15-
Kimi K2.5Moonshot AI, released 2026-01-27$1198.46Z.ai, GLM-5 model cardchecked 2026-10-15; printed on 1 more page-
GLM-5.1Z.ai, released 2026-04-07$5634.41Z.ai, GLM-5.1 model cardchecked 2026-10-15-
GLM-5Z.ai, released 2026-02-12$4432.12Z.ai, GLM-5 model cardchecked 2026-10-15; printed on 1 more page-
GLM-4.7Z.ai, released 2025-12-22$2376.82Z.ai, GLM-5 model cardchecked 2026-10-15-

Every number here is read from the same catalog file as the benchmark table, data version 2026103001, and is the number the table shows for that model and test. Where one page printed a model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.