Model Promotion Observatory // Phase 4.5

Why this model.
For this capability.

Inspect two accepted public certification records for the same job. Start with the capability, compare only represented evidence, then follow evaluation, human review, and promotion as separate gates.

Open Evidence Console Systems directory

This Observatory is read-only evidence. It is not a global score, benchmark page, deployment dashboard, live runtime monitor, or model-control interface.

01 // Capability first

Choose the job, then inspect both records.

No combined score is calculated. Evidence differences stay attached to the capability that produced them, and missing evidence stays UNKNOWN / NOT OBSERVED.

JavaScript is optional. All five comparison chambers remain in the document source; the selector only focuses the reading.

02 // Comparison chamber

Ramone RAG generation

ramone-rag-generation

Which evidence supports this job?

Candidate A // Accepted public receipt

qwen3.5-mtp

Bound identityqwen3.5-mtpCapability + fingerprint bound

Evidence result

3 / 3 passedRequired threshold 3 / 3

This candidate met the required threshold for ramone-rag-generation. Suitability remains specific to this capability and evidence set.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-15 22:30:46.620278 UTC
  2. EVALUATION PASSED3 / 3 passed · required threshold 3 / 3
  3. HUMAN REVIEWEDObserved · 2026-09-16 13:44:18 UTC
  4. PROMOTION APPROVEDObserved · 2026-09-16 13:44:18 UTC
Evaluation
Evaluation passed
Human review
Human reviewed
Promotion
Promotion approved
Categories
  • Abstention
  • Fabrication
  • Grounding
Known regressions
None recorded
  • None recorded
Freshness
Current evidence
Fingerprint
sha256:b92b3faa8d75a6869dbf72b0b9cf80c214cba272e23b3a12f859ed32c4deeba5
Suite
1.0.0
suite:sha256:bdac9a242fbec9eeef8fa3799aa73deedae0188843ceeb3ddfa7b34967e9f4e9
Observed
Runtime identity
Not represented in this receipt

Candidate B // Accepted public receipt

qwen3:14b

Bound identityqwen3:14bCapability + fingerprint bound

Evidence result

3 / 3 passedRequired threshold 3 / 3

This candidate met the required threshold for ramone-rag-generation. Suitability remains specific to this capability and evidence set.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-22 01:10:21.877434 UTC
  2. EVALUATION PASSED3 / 3 passed · required threshold 3 / 3
  3. HUMAN REVIEW PENDINGNo accepted review recorded
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation passed
Human review
Human review pending
Promotion
Promotion not approved
Categories
  • Abstention
  • Fabrication
  • Grounding
Known regressions
UNKNOWN / NOT OBSERVED
  • UNKNOWN / NOT OBSERVED
Freshness
Current evidence
Fingerprint
sha256:78eac8ff3e2e0aedc4ad6c8442c9d17d2d73890f06e914481d6d4e577ed9d83f
Suite
1.0.0
suite:sha256:bdac9a242fbec9eeef8fa3799aa73deedae0188843ceeb3ddfa7b34967f9e4e9
Observed
Runtime identity
Not represented in this receipt

Both records show 3 / 3 passed. Their lifecycle states differ: qwen3.5-mtp has an accepted promotion decision for this capability; qwen3:14b remains review pending and not approved. This does not establish deployment or current routing.

Preserved RAG evidence lineage

Historical qwen3.5 receipt · review pendingsha256:f80ade07d8025fa751f59f8e667fa1adb2905d194fa1f1c90fc316730a94b3c3
Current qwen3.5 receipt · promotion approvedHistorical evidence remains preserved.

Supersession changes evidence lineage, not deployment or runtime state.

02 // Comparison chamber

Ramone live chat

ramone-live-chat

Which evidence supports this job?

Candidate A // Accepted public receipt

qwen3.5-mtp

Bound identityqwen3.5-mtpCapability + fingerprint bound

Evidence result

3 / 3 passedRequired threshold 3 / 3

This candidate met the required threshold for ramone-live-chat. Suitability remains specific to this capability and evidence set.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-16 16:52:06.275921 UTC
  2. EVALUATION PASSED3 / 3 passed · required threshold 3 / 3
  3. HUMAN REVIEW PENDINGNo accepted review recorded
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation passed
Human review
Human review pending
Promotion
Promotion not approved
Categories
  • Abstention
  • Grounding
Known regressions
UNKNOWN / NOT OBSERVED
  • UNKNOWN / NOT OBSERVED
Freshness
Current evidence
Fingerprint
sha256:9d8d70d7577eb4e307852ca5b552370d02754d5d27205097b1393277a61c074e
Suite
1.0.0
suite:sha256:a06cbe278ac92d58e4588fa464642608fd55f6c068dd07b2a5cee0c116b32f90
Observed
Runtime identity
Not represented in this receipt

Candidate B // Accepted public receipt

qwen3:14b

Bound identityqwen3:14bCapability + fingerprint bound

Evidence result

3 / 3 passedRequired threshold 3 / 3

This candidate met the required threshold for ramone-live-chat. Suitability remains specific to this capability and evidence set.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-22 01:10:21.877434 UTC
  2. EVALUATION PASSED3 / 3 passed · required threshold 3 / 3
  3. HUMAN REVIEW PENDINGNo accepted review recorded
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation passed
Human review
Human review pending
Promotion
Promotion not approved
Categories
  • Abstention
  • Grounding
Known regressions
UNKNOWN / NOT OBSERVED
  • UNKNOWN / NOT OBSERVED
Freshness
Current evidence
Fingerprint
sha256:ae8fd23f04ae11f8d617aabc981a955180384e06f901d090bf64d7eed799e665
Suite
1.0.0
suite:sha256:a06cbe278ac92d58e4588fa464642608fd55f6c068dd07b2a5cee0c116b32f90
Observed
Runtime identity
Not represented in this receipt

02 // Comparison chamber

Corpus retrieval

corpus-retrieval

Which evidence supports this job?

Candidate A // Accepted public receipt

qwen3.5-mtp

Bound identityqwen3.5-mtpCapability + fingerprint bound

Evidence result

5 / 5 passedRequired threshold 5 / 5

This candidate met the required threshold for corpus-retrieval. Suitability remains specific to this capability and evidence set.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-16 16:52:06.275921 UTC
  2. EVALUATION PASSED5 / 5 passed · required threshold 5 / 5
  3. HUMAN REVIEW PENDINGNo accepted review recorded
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation passed
Human review
Human review pending
Promotion
Promotion not approved
Categories
  • Abstention
  • Causal claim
  • Grounding
Known regressions
UNKNOWN / NOT OBSERVED
  • UNKNOWN / NOT OBSERVED
Freshness
Current evidence
Fingerprint
sha256:8b6ffc407bc02af7cc01596c99856dba03112c1345d2875feb587fa3c87503aa
Suite
1.0.0
suite:sha256:78afced4d26777467be938a3f2c074946b025e5ee97973fb7e0b3722a14b77da
Observed
Runtime identity
Not represented in this receipt

Candidate B // Accepted public receipt

qwen3:14b

Bound identityqwen3:14bCapability + fingerprint bound

Evidence result

5 / 5 passedRequired threshold 5 / 5

This candidate met the required threshold for corpus-retrieval. Suitability remains specific to this capability and evidence set.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-22 01:10:21.877434 UTC
  2. EVALUATION PASSED5 / 5 passed · required threshold 5 / 5
  3. HUMAN REVIEW PENDINGNo accepted review recorded
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation passed
Human review
Human review pending
Promotion
Promotion not approved
Categories
  • Abstention
  • Causal claim
  • Grounding
Known regressions
UNKNOWN / NOT OBSERVED
  • UNKNOWN / NOT OBSERVED
Freshness
Current evidence
Fingerprint
sha256:79c2b00ce3c3ab2c6759481f63b1a3bf9b6eef1194bd72d715f0b40a318d79c3
Suite
1.0.0
suite:sha256:78afced4d26777467be938a3f2c074946b025e5ee97973fb7e0b3722a14b77da
Observed
Runtime identity
Not represented in this receipt

02 // Comparison chamber

Daily Digest synthesis

daily-digest-synthesis

Which evidence supports this job?

Candidate A // Accepted public receipt

qwen3.5-mtp

Bound identityqwen3.5-mtpCapability + fingerprint bound

Evidence result

2 / 3 passedRequired threshold 3 / 3

This candidate did not meet the required threshold for daily-digest-synthesis. The result is an evaluation failure, not a service failure.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-16 16:52:06.275921 UTC
  2. EVALUATION FAILED2 / 3 passed · required threshold 3 / 3
  3. HUMAN REVIEW NOT REQUIREDDownstream promotion is not approved
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation failed
Human review
Human review not required
Promotion
Promotion not approved
Categories
  • Abstention
  • Grounding
Known regressions
UNKNOWN / NOT OBSERVED
  • UNKNOWN / NOT OBSERVED
Freshness
Current evidence
Fingerprint
sha256:3e1dbfba5a3c9c643791b91b608810651f42212a019962275895f25eba21bdf3
Suite
1.0.0
suite:sha256:c0f96b4d820250a97104478fc072db0e7096d408525730715d47d054df93067a
Observed
Runtime identity
Not represented in this receipt

Candidate B // Accepted public receipt

qwen3:14b

Bound identityqwen3:14bCapability + fingerprint bound

Evidence result

1 / 3 passedRequired threshold 3 / 3

This candidate did not meet the required threshold for daily-digest-synthesis. The result is an evaluation failure, not a service failure.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-22 01:10:21.877434 UTC
  2. EVALUATION FAILED1 / 3 passed · required threshold 3 / 3
  3. HUMAN REVIEW NOT REQUIREDDownstream promotion is not approved
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation failed
Human review
Human review not required
Promotion
Promotion not approved
Categories
  • Abstention
  • Grounding
Known regressions
UNKNOWN / NOT OBSERVED
  • UNKNOWN / NOT OBSERVED
Freshness
Current evidence
Fingerprint
sha256:9756936d47899afab0a2718ef4eb34d2a0239cb56da3675ad19a368e1950cb14
Suite
1.0.0
suite:sha256:c0f96b4d820250a97104478fc072db0e7096d408525730715d47d054df93067a
Observed
Runtime identity
Not represented in this receipt

The accepted records show 2 / 3 versus 1 / 3, and both remain evaluated-failed. This is capability evidence, not a combined score or a universal model decision.

02 // Comparison chamber

Postmortem drafting

postmortem-drafting

Which evidence supports this job?

Candidate A // Accepted public receipt

qwen3.5-mtp

Bound identityqwen3.5-mtpCapability + fingerprint bound

Evidence result

6 / 7 passedRequired threshold 7 / 7

This candidate did not meet the required threshold for postmortem-drafting. The result is an evaluation failure, not a service failure.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-16 16:52:06.275921 UTC
  2. EVALUATION FAILED6 / 7 passed · required threshold 7 / 7
  3. HUMAN REVIEW NOT REQUIREDDownstream promotion is not approved
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation failed
Human review
Human review not required
Promotion
Promotion not approved
Categories
  • Abstention
  • Causal claim
  • Fabrication
  • Grounding
Known regressions
UNKNOWN / NOT OBSERVED
  • UNKNOWN / NOT OBSERVED
Freshness
Current evidence
Fingerprint
sha256:b049801c77b15a8efccf7a59bac0aaf41bae63f12bd4bb460bf5de662223053f
Suite
1.0.0
suite:sha256:f20ccf4ef9ccd714c8a751a4888bf05ba967f8daa454b406773b82d318ebab3b
Observed
Runtime identity
Not represented in this receipt

Candidate B // Accepted public receipt

qwen3:14b

Bound identityqwen3:14bCapability + fingerprint bound

Evidence result

6 / 7 passedRequired threshold 7 / 7

This candidate did not meet the required threshold for postmortem-drafting. The result is an evaluation failure, not a service failure.

Promotion trail

  1. EVAL PREPAREDObserved · 2026-09-22 01:10:21.877434 UTC
  2. EVALUATION FAILED6 / 7 passed · required threshold 7 / 7
  3. HUMAN REVIEW NOT REQUIREDDownstream promotion is not approved
  4. PROMOTION NOT APPROVEDNo approved promotion record
Evaluation
Evaluation failed
Human review
Human review not required
Promotion
Promotion not approved
Categories
  • Abstention
  • Causal claim
  • Fabrication
  • Grounding
Known regressions
Known regression
  • Format contract failure
Freshness
Current evidence
Fingerprint
sha256:acfce7f868020a9c0b633b92494f1825bcd52f2eddb1f429481da56ecf69837a
Suite
1.0.0
suite:sha256:f20ccf4ef9ccd714c8a751a4888bf05ba967f8daa454b406773b82d318ebab3b
Observed
Runtime identity
Not represented in this receipt

Both records show 6 / 7 passed and remain evaluated-failed. The qwen3:14b receipt additionally records a public format-contract-failure regression category. No private output, prompt, context, or scorer text is reproduced.

Hard authority boundary

PROMOTION APPROVED ≠ DEPLOYED

qwen3.5-mtp is promotion-approved only for ramone-rag-generation. qwen3:14b remains not approved for all five capabilities. These receipts do not prove installation, current routing, provider state, runtime health, or live verification.

03 // Regression Microscope

Inspect the bounded failure signature.

The microscope consumes only the accepted public fields known_regressions.state and known_regressions.categories. It never reconstructs a private output.

Expected behaviour is described at a safe abstraction level. UNKNOWN / NOT OBSERVED is not the same as None recorded. If a future accepted receipt is NOT APPLICABLE, that state is shown explicitly.

Ramone RAG generation // Candidate A

qwen3.5-mtp

None recorded
Failure signature
None recorded
Expected behaviour
The evaluated output should satisfy the declared capability contract.
Observed failure class
No public regression category recorded in this accepted projection.
State
None recorded

Ramone RAG generation // Candidate B

qwen3:14b

UNKNOWN / NOT OBSERVED
Failure signature
UNKNOWN / NOT OBSERVED
Expected behaviour
The evaluated output should satisfy the declared capability contract.
Observed failure class
The accepted projection does not establish a public regression category.
State
UNKNOWN / NOT OBSERVED

Ramone live chat // Candidate A

qwen3.5-mtp

UNKNOWN / NOT OBSERVED
Failure signature
UNKNOWN / NOT OBSERVED
Expected behaviour
The evaluated output should satisfy the declared capability contract.
Observed failure class
The accepted projection does not establish a public regression category.
State
UNKNOWN / NOT OBSERVED

Ramone live chat // Candidate B

qwen3:14b

UNKNOWN / NOT OBSERVED
Failure signature
UNKNOWN / NOT OBSERVED
Expected behaviour
The evaluated output should satisfy the declared capability contract.
Observed failure class
The accepted projection does not establish a public regression category.
State
UNKNOWN / NOT OBSERVED

Corpus retrieval // Candidate A

qwen3.5-mtp

UNKNOWN / NOT OBSERVED
Failure signature
UNKNOWN / NOT OBSERVED
Expected behaviour
The evaluated output should satisfy the declared capability contract.
Observed failure class
The accepted projection does not establish a public regression category.
State
UNKNOWN / NOT OBSERVED

Corpus retrieval // Candidate B

qwen3:14b

UNKNOWN / NOT OBSERVED
Failure signature
UNKNOWN / NOT OBSERVED
Expected behaviour
The evaluated output should satisfy the declared capability contract.
Observed failure class
The accepted projection does not establish a public regression category.
State
UNKNOWN / NOT OBSERVED

Daily Digest synthesis // Candidate A

qwen3.5-mtp

UNKNOWN / NOT OBSERVED
Failure signature
UNKNOWN / NOT OBSERVED
Expected behaviour
The evaluated output should satisfy the declared capability contract.
Observed failure class
The accepted projection does not establish a public regression category.
State
UNKNOWN / NOT OBSERVED

Daily Digest synthesis // Candidate B

qwen3:14b

UNKNOWN / NOT OBSERVED
Failure signature
UNKNOWN / NOT OBSERVED
Expected behaviour
The evaluated output should satisfy the declared capability contract.
Observed failure class
The accepted projection does not establish a public regression category.
State
UNKNOWN / NOT OBSERVED

Postmortem drafting // Candidate A

qwen3.5-mtp

UNKNOWN / NOT OBSERVED
Failure signature
UNKNOWN / NOT OBSERVED
Expected behaviour
The evaluated output should satisfy the declared presentation and output contract.
Observed failure class
The accepted projection does not establish a public regression category.
State
UNKNOWN / NOT OBSERVED

Postmortem drafting // Candidate B

qwen3:14b

Known regression
Failure signature
Format contract failure
Expected behaviour
The evaluated output should satisfy the declared presentation and output contract.
Observed failure class
The accepted evaluation recorded a format-contract violation.
State
Known regression

04 // Evidence limit

Performance timing is not compared here.

The accepted public receipts do not establish a comparable runtime protocol. No latency, tokens-per-second, speed winner, or private timing evidence is shown.