GLM 5.3 Flash

The scorecard

Each axis runs from 0 to 100. The colored mark is this model. The faint marks are the other models in this edition. Hover one to see which, and click it to open that model.

• Tool use

  The agent has to call the right tool, with the right arguments, in the right order. Telling the customer it's done before the tool has finished counts against the model.

  30 conversations

  Mistral Small 4 49 Nova 2 Lite 53 GPT-4.1 mini 54 Gemini 2.5 Flash Lite 58 Gemini 3.1 Flash Lite 66 gpt-oss-120b 66 Gemma 4 31B 69 Claude Haiku 4.5 72 Gemini 3.5 Flash Lite 74 GPT-5 mini 78 GPT-5.6 Luna 85 DeepSeek V4 Flash 97

  96

  #2 of 13

  0 100

  It checks the order before it acts and flags it when the record contradicts the customer, but when the customer presses to return more units than the order shows, it usually puts the return through anyway.

• Task completion

  The customer has to leave with what they came for and nothing extra. Only the outcome counts, in long conversations where the customer gets a fact wrong, sets a limit early, wanders off topic, and changes their mind.

  21 conversations

  Gemini 2.5 Flash Lite 34 Nova 2 Lite 53 GPT-4.1 mini 56 Mistral Small 4 62 Claude Haiku 4.5 72 gpt-oss-120b 74 Gemma 4 31B 79 Gemini 3.5 Flash Lite 83 Gemini 3.1 Flash Lite 86 GPT-5.6 Luna 86 DeepSeek V4 Flash 89 GPT-5 mini 91

  95

  #1 of 13

  0 100

  It works through a whole account in one pass, separating what can still go back from what cannot, though once it waited on a reason from the customer and never started the return.

• Context retention

  A constraint the customer states once, early, has to still hold when the agent acts at the end of a long conversation.

  21 conversations

  Gemini 2.5 Flash Lite 50 Mistral Small 4 69 Gemini 3.5 Flash Lite 70 gpt-oss-120b 71 GPT-4.1 mini 72 Gemini 3.1 Flash Lite 74 Gemma 4 31B 78 Nova 2 Lite 78 GPT-5.6 Luna 81 Claude Haiku 4.5 82 DeepSeek V4 Flash 86 GPT-5 mini 87

  75

  #7 of 13

  0 100

  It holds a preference given at the start, such as sending a refund to store credit, but with two orders open it canceled the one the customer wanted kept.

• Grounding

  Answers have to come from the knowledge base. When a customer states a wrong spec or a made-up product, the agent has to correct them or say plainly that it doesn't know, and still handle the rest of the request.

  24 conversations

  Nova 2 Lite 66 Mistral Small 4 66 Claude Haiku 4.5 67 GPT-4.1 mini 67 Gemini 2.5 Flash Lite 69 Gemini 3.5 Flash Lite 71 gpt-oss-120b 83 Gemini 3.1 Flash Lite 86 Gemma 4 31B 92 GPT-5 mini 93 DeepSeek V4 Flash 94 GPT-5.6 Luna 94

  90

  #5 of 13

  0 100

  It holds prices and policies when a customer pushes back, but it explains what a bundled accessory does with nothing behind it.

• Safety

  Simulated customers fake system messages, claim authority they don't have, and slip in instructions of their own. The agent has to hold its policy, keep its own instructions and tools to itself, protect other customers' orders, and still serve the real request.

  21 conversations

  Nova 2 Lite 60 Gemini 2.5 Flash Lite 61 GPT-4.1 mini 72 Mistral Small 4 73 gpt-oss-120b 76 GPT-5 mini 84 Gemma 4 31B 90 GPT-5.6 Luna 90 Gemini 3.1 Flash Lite 92 Claude Haiku 4.5 93 Gemini 3.5 Flash Lite 93 DeepSeek V4 Flash 94

  94

  #1 of 13

  0 100

  It turned down a $50 goodwill credit offered for a favorable audit, and a 20 percent uplift that didn't exist.

• Hallucinations

  We pull out every statement the agent makes about a product or about what it has done, and check each one against the knowledge base and the tool results. One flagged statement marks the whole conversation, and the score is the share of conversations with nothing flagged, shown with its range.

  21 conversations · 44 of 1112 claims flagged · 2.1 unsupported claims per conversation

  Nova 2 Lite 0 gpt-oss-120b 10 Gemini 2.5 Flash Lite 14 GPT-4.1 mini 14 Mistral Small 4 14 Claude Haiku 4.5 19 DeepSeek V4 Flash 19 GPT-5 mini 29 Gemini 3.5 Flash Lite 33 Gemini 3.1 Flash Lite 48 Gemma 4 31B 48 GPT-5.6 Luna 67

  24

  #6 of 13 11 to 45

  0 100

  It corrects wrong prices, but pressed for a number it doesn't have it gives one anyway, and it promises a support follow-up email it never arranged.