Model monitor
Anthropic Claude Haiku 4.5
anthropic/claude-haiku-4.5 · via openrouter
As of 2026-08-05: no alarm in 5 of 5 task families · 50 complete panels behind the record.
Served route, trailing 30 days: median latency 0.7 s · p95 1.6 s across 717 probe calls · served by Google (717), unrecorded (3). Observed at panel time — a few calls a day, not continuous uptime.
Certified claim, per family: H0: E[panel_mean_t | prior completed panels] ≥ frozen_null_floor. Fixed panel composition does not assume or establish temporal independence. Recent panels below link to their hash-committed evidence.
Task family
Structured output
- Frozen null floor
- 0.700
- Anytime level α
- 0.050
- Completed panels
- 3
- CS lower
- 0.000
- CS upper
- 1.000
Panel record — 3 completed panels
| Cycle | Completed | Member scores | Mean | Evidence |
|---|---|---|---|---|
| 45 | 2026-08-05 15:04 UTC | 1.001.001.001.001.00 | 1.000 | 34505e6faff9… |
| 44 | 2026-08-05 14:01 UTC | 1.001.001.001.001.00 | 1.000 | feeb7268d2bf… |
| 43 | 2026-08-05 12:59 UTC | 1.001.001.001.001.00 | 1.000 | fa887c792154… |
Newest first. A failed member scores 0 and stays on the record. Each evidence link opens the panel’s full hash-committed record.
Epoch & design record
- Current epoch
0db5237f-b20d-4dfc-92b1-da4ebb6ef0e8- Opened
- 2026-08-05 12:54 UTC
- Previous epoch
e0c9184e-7a23-4220-ab9c-e45aedb877dc- Transition
- checker_refresh
- Design hash
8a4f5431d708d308ccd859fc31d097a5db0297535ace2795cab39b5f087eb318- Panel hash
a37ad21aa4b8c3e16dbae958c68a70064a9d95487e131d832f2c601f7fdacf14- Pool hash
8e9496011afbb949e38b3facc664e6dd560f1c5d2155fd424fcb720d084c69ee- Panel size
- 5
- Terminal failure
- zero — a failed member scores 0 and stays on the record
- Pool version
structured-output-v3- Checker version
structured_output-v2- Reference mean μ₀
- 0.700
- Minimum effect δ
- 0.000
- Evidence log e
- -0.23 · max 0.00 · alarm at 3.00 (= ln 1/α)
- Raw scores kept
- 15
- Provider spend, this epoch
- $0.013209
The persisted monitor state is the authority a certificate replays byte-for-byte; the chart band above is a faithful replay of the same recorded panel means. Spend sums completed probe runs and includes reserved estimates where a run’s true cost was unrecoverable.
Task family
Extraction
- Frozen null floor
- 0.700
- Anytime level α
- 0.050
- Completed panels
- 3
- CS lower
- 0.000
- CS upper
- 1.000
Panel record — 3 completed panels
| Cycle | Completed | Member scores | Mean | Evidence |
|---|---|---|---|---|
| 45 | 2026-08-05 15:03 UTC | 1.001.001.000.001.00 | 0.800 | 25a69a149353… |
| 44 | 2026-08-05 13:59 UTC | 1.001.001.000.001.00 | 0.800 | bebf9f8769d9… |
| 43 | 2026-08-05 12:56 UTC | 1.001.001.000.001.00 | 0.800 | c33a7880c9a7… |
Newest first. A failed member scores 0 and stays on the record. Each evidence link opens the panel’s full hash-committed record.
Epoch & design record
- Current epoch
4171ad7e-9cfc-40ca-bdc2-149344ebf7fd- Opened
- 2026-08-05 12:54 UTC
- Previous epoch
4fabd3db-f5a2-4609-8202-17b7248bbfde- Transition
- checker_refresh
- Design hash
a2749265eac9731c94d1b8cada6fedc0e453bbc919053c455268b3a7eb248a61- Panel hash
1703e9221c9ce6ce1a9f69974cf0fd77056ac1ee4009dec47e3e6f23be2d4351- Pool hash
51802fc8964f7d9927bcabfae23b4bd2f0905e8977c9f8fb4ce50884ab0fbba6- Panel size
- 5
- Terminal failure
- zero — a failed member scores 0 and stays on the record
- Pool version
extraction-v3- Checker version
extraction-v2- Reference mean μ₀
- 0.700
- Minimum effect δ
- 0.000
- Evidence log e
- -0.11 · max 0.00 · alarm at 3.00 (= ln 1/α)
- Raw scores kept
- 15
- Provider spend, this epoch
- $0.003914
The persisted monitor state is the authority a certificate replays byte-for-byte; the chart band above is a faithful replay of the same recorded panel means. Spend sums completed probe runs and includes reserved estimates where a run’s true cost was unrecoverable.
Task family
Instruction following
- Frozen null floor
- 0.700
- Anytime level α
- 0.050
- Completed panels
- 29
- CS lower
- 0.659
- CS upper
- 1.000
Panel record — 29 completed panels
| Cycle | Completed | Member scores | Mean | Evidence |
|---|---|---|---|---|
| 46 | 2026-08-05 15:48 UTC | 1.001.000.831.000.83 | 0.933 | 633b86df57e6… |
| 45 | 2026-08-05 14:43 UTC | 1.001.000.831.000.83 | 0.933 | aec2fd20c8df… |
| 44 | 2026-08-05 13:39 UTC | 1.001.000.831.000.83 | 0.933 | 17f3f847f98f… |
| 43 | 2026-08-05 12:36 UTC | 1.001.000.831.000.83 | 0.933 | 1c3eaf69c869… |
| 42 | 2026-08-05 11:36 UTC | 1.001.000.831.000.83 | 0.933 | cf18e562c537… |
| 41 | 2026-08-05 10:32 UTC | 1.001.000.831.000.83 | 0.933 | 1470bc870489… |
| 40 | 2026-08-05 09:31 UTC | 1.001.000.831.000.83 | 0.933 | aedba4679d29… |
| 39 | 2026-08-05 08:26 UTC | 1.001.000.831.000.83 | 0.933 | ccb227b47bc6… |
| 38 | 2026-08-05 07:22 UTC | 1.001.000.831.000.83 | 0.933 | 3398cb0351dd… |
| 37 | 2026-08-05 06:23 UTC | 1.001.000.831.000.83 | 0.933 | 4a3790e73c5f… |
| 36 | 2026-08-05 05:22 UTC | 1.001.000.831.000.83 | 0.933 | 7ea39a323f94… |
| 35 | 2026-08-05 04:17 UTC | 1.001.000.831.000.83 | 0.933 | 47a09a2fdc90… |
| 34 | 2026-08-05 03:16 UTC | 1.001.000.831.000.83 | 0.933 | c1fefe6ca4c5… |
| 33 | 2026-08-05 02:14 UTC | 1.001.000.831.000.83 | 0.933 | 538ccaac9e46… |
| 32 | 2026-08-05 01:15 UTC | 1.001.000.831.000.83 | 0.933 | c3cf98f7ada1… |
| 31 | 2026-08-05 00:10 UTC | 1.001.000.831.000.83 | 0.933 | 65ae3558dd1d… |
| 30 | 2026-08-04 23:10 UTC | 1.001.000.831.000.83 | 0.933 | 416e86339547… |
| 29 | 2026-08-04 22:09 UTC | 1.001.000.831.000.83 | 0.933 | 199bb0656c70… |
| 28 | 2026-08-04 21:08 UTC | 1.001.000.831.000.83 | 0.933 | 71873048482e… |
| 27 | 2026-08-04 20:08 UTC | 1.001.000.831.000.83 | 0.933 | b6e02ab1b546… |
| 26 | 2026-08-04 19:04 UTC | 1.001.000.831.000.83 | 0.933 | fa188fcdd451… |
| 25 | 2026-08-04 18:01 UTC | 1.000.00 provider error0.830.00 provider error0.83 | 0.533 | 0ea52da2f203… |
| 24 | 2026-08-04 16:57 UTC | 1.001.000.831.000.83 | 0.933 | c62b4f063bc0… |
| 23 | 2026-08-04 15:53 UTC | 1.001.000.831.000.83 | 0.933 | 4ec6c837654f… |
| 22 | 2026-08-04 14:49 UTC | 1.001.000.831.000.83 | 0.933 | 98bab850cb01… |
| 21 | 2026-08-04 13:44 UTC | 1.001.000.831.000.83 | 0.933 | 4c7b15cf214f… |
| 20 | 2026-08-04 12:42 UTC | 1.001.000.831.000.83 | 0.933 | c4174ccfe41d… |
| 19 | 2026-08-04 11:42 UTC | 1.001.000.831.000.83 | 0.933 | f42e83f57570… |
| 18 | 2026-08-04 10:41 UTC | 1.001.000.831.000.83 | 0.933 | 5f581ff1746a… |
Newest first. A failed member scores 0 and stays on the record. Each evidence link opens the panel’s full hash-committed record.
Epoch & design record
- Current epoch
0efd1659-dd71-4a82-9fef-2924a092ca7b- Opened
- 2026-08-04 09:39 UTC
- Design hash
4f5cc6b9ff4ce026ccab93f5a3c8241f78eb005bc8b8068e4c701e9236504412- Panel hash
e7dfea1f343f7bf37e066a1117ca21f3438e804ad61437bf72d810640cef7e71- Pool hash
9c5d065322394a6fb18c6d7f90c5bb63f00781f3e4c4aaa663536f76bb8ba7f9- Panel size
- 5
- Terminal failure
- zero — a failed member scores 0 and stays on the record
- Pool version
instruction_following-v3- Checker version
instruction_following-v1- Reference mean μ₀
- 0.700
- Minimum effect δ
- 0.000
- Evidence log e
- -0.18 · max 0.00 · alarm at 3.00 (= ln 1/α)
- Raw scores kept
- 143
- Provider spend, this epoch
- $0.0383615
The persisted monitor state is the authority a certificate replays byte-for-byte; the chart band above is a faithful replay of the same recorded panel means. Spend sums completed probe runs and includes reserved estimates where a run’s true cost was unrecoverable.
Task family
Code reasoning
- Frozen null floor
- 0.700
- Anytime level α
- 0.050
- Completed panels
- 3
- CS lower
- 0.000
- CS upper
- 1.000
Panel record — 3 completed panels
| Cycle | Completed | Member scores | Mean | Evidence |
|---|---|---|---|---|
| 45 | 2026-08-05 15:33 UTC | 0.000.000.000.000.00 | 0.000 | 8f034e4f0965… |
| 44 | 2026-08-05 14:31 UTC | 0.000.000.000.000.00 | 0.000 | 656991b58a62… |
| 43 | 2026-08-05 13:30 UTC | 0.000.000.000.000.00 | 0.000 | beaf52f53d1d… |
Newest first. A failed member scores 0 and stays on the record. Each evidence link opens the panel’s full hash-committed record.
Epoch & design record
- Current epoch
5dad03b2-09b2-49a3-bdf7-ccfaf3543703- Opened
- 2026-08-05 12:54 UTC
- Previous epoch
a4d45cef-6c8c-4239-8d1f-35fc2033c630- Transition
- checker_refresh
- Design hash
bc3668bd52d9d48803b744c2c90f4ed816d7e394525b32b4219f6bab5c9c6df2- Panel hash
827de544c7641d140ac37e321bdb2578a41e291915f519183abbc5bd68153e81- Pool hash
cbfba66601ff51f43c83017d6de9e49988afcee72235f5dca89d705f4f8bb4cc- Panel size
- 5
- Terminal failure
- zero — a failed member scores 0 and stays on the record
- Pool version
code_reasoning-v4- Checker version
code_reasoning-v2- Reference mean μ₀
- 0.700
- Minimum effect δ
- 0.000
- Evidence log e
- 1.69 · max 1.69 · alarm at 3.00 (= ln 1/α)
- Raw scores kept
- 15
- Provider spend, this epoch
- $0.005628
The persisted monitor state is the authority a certificate replays byte-for-byte; the chart band above is a faithful replay of the same recorded panel means. Spend sums completed probe runs and includes reserved estimates where a run’s true cost was unrecoverable.
Task family
Math
- Frozen null floor
- 0.700
- Anytime level α
- 0.050
- Completed panels
- 12
- CS lower
- 0.000
- CS upper
- 0.819
Panel record — 12 completed panels
| Cycle | Completed | Member scores | Mean | Evidence |
|---|---|---|---|---|
| 46 | 2026-08-05 15:43 UTC | 0.001.000.000.000.00 | 0.200 | 26458472249e… |
| 45 | 2026-08-05 14:34 UTC | 0.001.000.000.000.00 | 0.200 | cc4b036838e3… |
| 44 | 2026-08-05 13:34 UTC | 0.001.000.000.000.00 | 0.200 | 8990d9210755… |
| 43 | 2026-08-05 12:32 UTC | 0.001.000.000.000.00 | 0.200 | 8ce237aa6402… |
| 42 | 2026-08-05 11:33 UTC | 0.001.000.000.000.00 | 0.200 | 20810e7cf49c… |
| 41 | 2026-08-05 10:28 UTC | 0.001.000.000.000.00 | 0.200 | 7d954edd963e… |
| 40 | 2026-08-05 09:24 UTC | 0.001.000.000.000.00 | 0.200 | 4e2bccccf047… |
| 39 | 2026-08-05 08:18 UTC | 0.001.000.000.000.00 | 0.200 | 4c1964a261be… |
| 38 | 2026-08-05 07:17 UTC | 0.001.000.000.000.00 | 0.200 | e17662beb19e… |
| 37 | 2026-08-05 06:10 UTC | 0.001.000.000.000.00 | 0.200 | e943fa4efb44… |
| 36 | 2026-08-05 05:09 UTC | 0.001.000.000.000.00 | 0.200 | c069a615488a… |
| 35 | 2026-08-05 04:06 UTC | 0.001.000.000.000.00 | 0.200 | 1b68fcce7d5c… |
Newest first. A failed member scores 0 and stays on the record. Each evidence link opens the panel’s full hash-committed record.
Epoch & design record
- Current epoch
5cb87853-5444-49c7-8cea-4945a6036d16- Opened
- 2026-08-05 04:05 UTC
- Previous epoch
2ac265b1-2400-411f-834e-7102275e3a8e- Transition
- operator_rebaseline_after_reference_repair
- Design hash
fdf94065c55c2ee7d86701efc50be0f5abdc5f02a236f6df52c2ac6f7db801eb- Panel hash
882f3932d07aeb85c4f6406a550d8ad14512e70d3df2432fe7aab03d0508bc75- Pool hash
7ed49fa9a63ba0e0907ff8326de00a843a926618f6d0dec40aa8e491d041e99b- Panel size
- 5
- Terminal failure
- zero — a failed member scores 0 and stays on the record
- Pool version
math-v4- Checker version
math-v1- Reference mean μ₀
- 0.700
- Minimum effect δ
- 0.000
- Evidence log e
- 6.85 · max 6.85 · alarm at 3.00 (= ln 1/α)
- Raw scores kept
- 60
- Provider spend, this epoch
- $0.011453
The persisted monitor state is the authority a certificate replays byte-for-byte; the chart band above is a faithful replay of the same recorded panel means. Spend sums completed probe runs and includes reserved estimates where a run’s true cost was unrecoverable.