OpenAI - notes on coverage

gtp-5-search-api

Including gtp-5-search-api-2025-10-14 snapshot

As of 2026-09-14

• This is the only remaining model that allows web-search via the v1/chat/completions endpoint.
• There is no model card at the expected location https://developers.openai.com/api/docs/models/gpt-5-search-api
• It is not listed on the deprecations page.
• It still answers API calls, but the response does not contain a tool usage object, so to calculate the cost you'd need to send us the web search count anyway.
• Seems Likely to be shut down soon.
• Not planning to support this.

Image generation

OpenAI offers image generation via

• Direct calls to image generation models via the v1/images/generations endpoint.
• Indirect calls via text models (from gpt-4.1 onwards, and only via the v1/responses endpoint) which can invoke them as tools.

Direct image generation models

• gpt-image-1 (2025-04)
• gpt-image-1-mini (2025-09)
• gpt-image-1.5 (2025-11)
• gpt-image-2 (2026-04)
• gpt-image-2.5-flare (2026-09) a faster, lower precision model
• gpt-image-2.5-sunburst (2026-09) a slower, higher precision model

If you prompt text models via the v1/chat/completions endpoint, you'll usually one of the following: an ASCII art image, a SVG image, instructions for producing a physical image.

Image editing

OpenAI offers image editing via its specialised image generation models called via the v1/images/edits endpoint.

• gpt-image-1 (2025-04)
• gpt-image-1-mini (2025-09)
• gpt-image-1.5 (2025-11)
• gpt-image-2 (2026-04)
• gpt-image-2.5-flare (2026-09) a faster, lower precision model
• gpt-image-2.5-sunburst (2026-09) a slower, higher precision model

Text models (from gpt-4.1 onwards) can invoke the image editing models as tools, when called via the v1/responses endpoint.

We haven't tested prompting text models for image editing via the v1/chat/completions endpoint.

Video generation

Video generation was available via the Sora models. These were shut down September 24, 2026.

/v1/chat/completions response shape

        {
          "object": "chat.completion",
          "model": "gpt-6-astra",
          "choices": [],
          "usage": {
              "prompt_tokens": 26,
              "completion_tokens": 337,
              "total_tokens": 363,
              "prompt_tokens_details": {
                  "cached_tokens": 0,
                  "cache_write_tokens": 0,
                  "audio_tokens": 0
              },
              "completion_tokens_details": {
                  "reasoning_tokens": 266,
                  "audio_tokens": 0,
                  "accepted_prediction_tokens": 0,
                  "rejected_prediction_tokens": 0
              }
          },
          "service_tier": "default",
        }
      

/v1/responses

        {
            "object": "response",
            "billing": {
                "payer": "openai"
            },
            "model": "gpt-5.6-sol",
            "output": [],
            "tool_usage": {
                "image_gen": {
                    "input_tokens": 66,
                    "input_tokens_details": {
                        "image_tokens": 0,
                        "text_tokens": 66
                    },
                    "output_tokens": 196,
                    "output_tokens_details": {
                        "image_tokens": 196,
                        "text_tokens": 0
                    },
                    "total_tokens": 262
                },
                "web_search": {
                    "num_requests": 1
                }
            },
            "tools": [
                {
                    "type": "code_interpreter",
                    "container": {
                        "type": "auto"
                    }
                },
                {
                    "type": "web_search"
                },
                {
                    "type": "image_generation",
                    "model": "gpt-image-2",
                    "n": 1,
                }
            ],
            "usage": {
                "input_tokens": 12148,
                "input_tokens_details": {
                    "cache_write_tokens": 5794,
                    "cached_tokens": 0
                },
                "output_tokens": 676,
                "output_tokens_details": {
                    "reasoning_tokens": 428
                },
                "total_tokens": 12824
            }
        }
        

/v1/images/generations

        {
          "data": [],
          "usage": {
              "input_tokens": 18,
              "input_tokens_details": {
                  "image_tokens": 0,
                  "text_tokens": 18
              },
              "output_tokens": 196,
              "output_tokens_details": {
                  "image_tokens": 196,
                  "text_tokens": 0
              },
              "total_tokens": 214
          }
        }
      

NOTE"model" and "object" (used for inferring endpoint) are both absent in this response.

/v1/images/edits

        {
          "data": [],
          "usage": {
              "input_tokens": 18,
              "input_tokens_details": {
                  "image_tokens": 0,
                  "text_tokens": 18
              },
              "output_tokens": 196,
              "output_tokens_details": {
                  "image_tokens": 196,
                  "text_tokens": 0
              },
              "total_tokens": 214
          }
        }
      

NOTE"model" and "object" (used for inferring endpoint) are both absent in this response.