Can GPT-5.6 terra search only allowed websites?
GPT-5.6 terra fails this allowed-sites search task. The outside-site page does not come back, but the answer also returns no cited page from the permitted site, so the required pair is incomplete.
1 of 1 test case failed
Every test case in this scenario has a result.
The test case
Each test case is judged on its own: Pass, Fail, Not gradable, or Untested. The scenario result above is based on this test case.
Search only these websitesFailEvidence
“In what year did GitHub first launch a technical preview of GitHub Copilot?”
- raw_response › output 4 › content 1 › text
- GitHub first launched GitHub Copilot as a **technical preview in 2021** (June 2021).
- raw_response › output 4 › content 1 › annotations
- []
- raw_response › output 1 › action › queries
- ["site:docs.github.com GitHub Copilot technical preview launch"]
- timestamp_utc
- 2026-09-28T05:44:04Z
Show the whole file (7 KB)
{ "run_id": "2026-09-28-r3-terra", "round": 2, "tool": "terra", "mode": "answer", "tc_id": "TC4", "capability": "C1", "scenario": "S4", "suite_role": "probe", "query_text": "In what year did GitHub first launch a technical preview of GitHub Copilot?", "restrict_domain": "docs.github.com", "restriction_mechanism": "native_param:filters.allowed_domains", "answer_key": "github.blog: 'Today, we are launching a technical preview of GitHub Copilot...' (June 29, 2021) -> 2021. docs.github.com page states no launch year.", "key_urls": [ { "url": "https://github.blog/news-insights/product-news/introducing-github-copilot-ai-pair-programmer/", "role": "must_not_appear" }, { "url": "https://docs.github.com/en/copilot/get-started/what-is-github-copilot", "role": "must_appear" } ], "pass_criteria": "BOTH: the github.blog answer page does NOT come back, AND a relevant docs.github.com page DOES", "key_verified_date": "2026-09-07", "before_check_rule": null, "timestamp_utc": "2026-09-28T05:44:04Z", "region": "ap-south-1", "status": "ok", "graded": null, "pass_fail": null, "grader_notes": null, "latency_ms": 6274.3, "http_status": 200, "error": null, "attempts": 1, "credits_reported": null, "tokens_reported": 19002, "results": [], "answer_text": "GitHub first launched GitHub Copilot as a **technical preview in 2021** (June 2021).", "citations": [], "raw_response": { "id": "resp_04cb92e032c2e7c1006ab9fe9f483087d19ce7744365ecbc4c", "object": "response", "created_at": 1790574239, "status": "completed", "access_programs": null, "background": false, "billing": { "payer": "developer" }, "completed_at": 1790574244, "error": null, "frequency_penalty": 0.0, "incomplete_details": null, "instructions": null, "max_output_tokens": null, "max_tool_calls": null, "model": "gpt-5.6-terra", "moderation": null, "output": [ { "id": "ws_04cb92e032c2e7c1006ab9fe9fe04c87d1a245b8bad8b5fee9", "type": "web_search_call", "status": "completed", "action": { "type": "search", "queries": ["site:docs.github.com GitHub Copilot technical preview launch"], "query": "site:docs.github.com GitHub Copilot technical preview launch" } }, { "id": "rs_04cb92e032c2e7c1006ab9fea12b1887d1ba462cd461da8705", "type": "reasoning", "content": [], "encrypted_content": "gAAAAABquf6kUpZ1y2uf7j72Egs5pefIsf9isNAy6L6g9rFTMfJpDlqNqNTekAm45raj-heHKM6nPFtyTGgxqAZ2qTAamUBfR7RQH6wF5vV3U7TnkiFSp7NsQh2CdRVZWbJWlrgjXYYMp4kfO5gQ7gDcY4Rv62zrPZsudzF1LwXpTbH4rJokUnEt1mGGa0GE1USRh54jjMrBslv8Ix8bwMfDwbUr9TpgqJqqDDs8FlBgP1knlDrUpVvxvL4RDZrZPN7kO9QOv4syqkzsQVZlWTUyVrJRpDTkXRP_YhYBVHAHM0QmihSbfEHhEFgD38jW3ETZg_HCJ4_rkYFphIFLpfKZ0HngxOoA_pJRXyvwzXAaVjEmbOlqj9_tu0D1EBg4VdVUyir05Q0nHvlaEsXDWvZbBPd8DGspCdAMfjVmAOW3nvsQHqJQ_kmOMQSp29IBU9OoRQIEDEzfQFLQOSNmsQwWx5nbFZtWqMRSsyBTaiFhXzDjkiR9wXVRtezp6DvLY-dMRzx2AqTUfprdA_bKQYKiWCiIKrlOpHH3pTJXvwtWUXdpCMJQgtAiFSKt9N4-P0CT8L7QQpHSsRTMAce8dMXXSYe05ruG0opmoiYK_vyLCa2eY2pC55WYnr90qfZBvz5va7vepEVp_Fn8t8zxiYL9cDfmvPLOakGzi3NjPMb98FTgqfBtj-66nQWdNiu-FfrHukyD5RAFe2F4ipaiWO7PpwzZ480zTU5qeyGCycESpsrnkwdsWeh59Uzek32xR13z0P8gGSaUZxpq7MBhl2ZWO8_W5MmAp5H7vA-SEgL_k_HxiiOganSGIzT60PWPjP8PriQ9KzKkqTVZlwj_sXKEPtgpOMIdb-0gnaVrexpRQ7L2xz0Kz1Qbw1Y2_-0aHmWQWz8UHgHwfJd1TOOHT8vModyNBA2SfIy_-VzPfXcjOY_Z10P_5-S4yM2FU6qa9bCC2LhwvacByirik5hB6TlWlDNFJsAPmWBaFpqJt_aTCklgENL7MXhz0tO11UuKwgEHErUEBIiBWcUc_Tzi4ur9-qtOp_s8FmK3J-fzq5DZ0QSJ9EEYk2E_LSHWZR3-kAulww-yaAOiog5jeUSD-xBxMAwS1xvm25ll0P1zkj-02crar9f45BCjjkLtISFmp2fJP009_wSJhf_wC5QssXvkGg0u_CTADmqo3dvkDqDuOzrjd6XXa6SxLNeVn1kdDP2j6A2fbf3zFlLj5aWRnUnwNWkq1HoSKe0WVFz0YI31FjiodkzPnE5ItFpkD_dyyAdM8-MK421HbCzrkJl5XtT7l8ye328xTOOt5LhfrENvvoWl0oHJ10URuyFrqZumXIlVKsz0xnOGfHeGnOj3q7YivKJUh64wobKNvJP6-vRUegWPrfdi0YCDmmbpei7GdxhvxGWE4fL3MKPQ0MXalAp4xn1bL-SJ58jpGdvv_R9tq9dVcEMLZfnX0bQFFkU_qhCNbtQvXnO_ZElL9FKN-YNCJT_Ibe4xtBDdKd2kq9EJxpetIfyEc_k=", "summary": [] }, { "id": "ws_04cb92e032c2e7c1006ab9fea1cb8c87d1ab52065e533b5640", "type": "web_search_call", "status": "completed", "action": { "type": "search", "queries": [ "site:docs.github.com \"technical preview\" \"GitHub Copilot\" 2021", "site:docs.github.com/en \"GitHub Copilot\" \"June 2021\"", "site:docs.github.com/en \"GitHub Copilot was\" \"technical preview\"" ], "query": "site:docs.github.com \"technical preview\" \"GitHub Copilot\" 2021" } }, { "id": "msg_04cb92e032c2e7c1006ab9fea38a2487d1a3e486358a97a954", "type": "message", "status": "completed", "content": [ { "type": "output_text", "annotations": [], "logprobs": [], "text": "GitHub first launched GitHub Copilot as a **technical preview in 2021** (June 2021)." } ], "phase": "final_answer", "role": "assistant" } ], "parallel_tool_calls": true, "presence_penalty": 0.0, "previous_response_id": null, "prompt_cache_key": null, "prompt_cache_retention": "24h", "reasoning": { "context": "all_turns", "effort": "medium", "mode": "standard", "summary": null }, "safety_identifier": null, "service_tier": "default", "store": true, "temperature": 1.0, "text": { "format": { "type": "text" }, "verbosity": "medium" }, "tool_choice": "auto", "tool_usage": { "image_gen": { "input_tokens": 0, "input_tokens_details": { "image_tokens": 0, "text_tokens": 0 }, "output_tokens": 0, "output_tokens_details": { "image_tokens": 0, "text_tokens": 0 }, "total_tokens": 0 }, "web_search": { "num_requests": 2 } }, "tools": [ { "type": "web_search", "filters": { "allowed_domains": ["docs.github.com"] }, "return_token_budget": "default", "search_content_types": ["text"], "search_context_size": "medium", "user_location": { "type": "approximate", "city": null, "country": "US", "region": null, "timezone": null } } ], "top_logprobs": 0, "top_p": 0.98, "truncation": "disabled", "usage": { "input_tokens": 18802, "input_tokens_details": { "cache_write_tokens": 4394, "cached_tokens": 0 }, "output_tokens": 200, "output_tokens_details": { "reasoning_tokens": 172 }, "total_tokens": 19002 }, "user": null, "metadata": {} } }
| ✗ | Expected: No page from the other site comes back, and at least one relevant page from docs.github.com does.Found: No github.blog page came back, but the answer cited no page at all, so no docs.github.com page came back either. |
The test needed both: no page from the other site, and at least one relevant page from docs.github.com. The first held. The answer cited no page at all, so the second did not, and one missing half makes this a Fail. The call itself completed normally.
Searches ran before the answer. Before answering, the model ran two searches with four queries, each limited to docs.github.com. Its answer still cited no page.
Launch year given without a source. The answer named 2021 as the launch year, a year the test places outside docs.github.com, and attached no source to it.
Configuration and setup
How this tool was set up for the run and what the test needed in place. Each row is a fact from the run's records; a fact the records do not hold is left out, not guessed.
How this scenario is graded
How we decide Pass, Fail and Not gradable. The same rules apply to every tool tested on this scenario.
Each test case gets one result per tool: Pass, Fail or Not gradable. A test case we haven't run yet shows Untested. There are no partial results.
- Pass: the tool did everything the test expected, and nothing it said contradicts the correct answer.
- Fail: at least one expected behaviour clearly didn't happen; the test case's result says which one and quotes the tool.
- Not gradable: our evidence couldn't settle the outcome (for example a record we needed is missing). It is never counted as a fail, and the test case's result says what's missing.
Where this sits in the benchmark
This page is one cell of a larger study: one tool, one scenario. Only this benchmark's frame appears here.
| Level | Name | Scope |
|---|---|---|
| Benchmark | Web Search for AI → | 10 scenarios · 12 tools |
| Capability | Web Retrieval → | |
| Scenario | Search only these websites → | |
| Tool | GPT-5.6 terra → |
Global scenario definition → · Global capability definition →
History of this result
What has happened to this result since it was first published. Runs and grades are never overwritten: a retest or a re-grade publishes a new result and keeps the earlier one readable.
Act on this result
Nothing filed here edits the run or the grade. A challenge opens a review, and a review can produce a new run or a re-grade — which becomes the current result and leaves this one in the history.
You run the same kind of test against your own setup and get the same behaviour.
Agree →Yours behaves differently. Tell us what you got, with a screenshot if you have one.
Disagree →Something here is wrong — a reference value, a transcription, a grade.
Report an issue →Tell us what changed and we schedule a rerun of the failing test case. The old result stays as history.
Vendor notice →The same record is available as structured data through the AI Demos MCP server, with the counts, the coverage and every per-test-case reason carried as fields.
Verify the proof files
These files support this result. Open a file to inspect the original evidence.
File fingerprints (SHA-256)
A fingerprint identifies the exact file used for this result.