Last refresh: 2026-05-17 · 17 models graded 3 OPEN RUBRIC ITEMS · DeepSeek judge coverage: 98%

Ten sample questions

A representative slice of the benchmark — two MCQs, two scenarios, one standards-conflict, two spot-the-error artefacts, one quantitative uncertainty problem, and two Part-C methodology questions. The full rubric for all 111 questions, every model's verdict on each, and the methodology paper text are in the full report.

10 of 111 questions spanning all 5 Part-A types and Part C CC-BY 4.0

A2 Biogenic carbon sub-indicators under EN 15804+A2 MCQ
Domain: construction | Standard: EN 15804+A2 | Difficulty: 1 | Verification: TRAINING-HIGH

Which GWP sub-indicators must be separately reported under EN 15804+A2?

A) GWP-total only. B) GWP-fossil and GWP-biogenic. C) GWP-fossil, GWP-biogenic, GWP-luluc, and GWP-total. D) GWP-fossil, GWP-biogenic, GWP-luluc, GWP-total, and GWP-aviation.

Correct: C. EN 15804+A2 requires disaggregation into fossil, biogenic, and land-use-and-land-use-change contributions, plus the total. All four must appear even if values are zero.

---

A9 PEF Circular Footprint Formula A coefficient range MCQ
Domain: packaging | Standard: PEF method, CFF | Difficulty: 2 | Verification: VERIFIED-W

In the PEF Circular Footprint Formula, the allocation factor A distributes burdens and credits between the supplier and user of recycled material. Its range is:

A) 0 to 1, continuous. B) 0.2 to 0.8, with specified defaults per material. C) Exactly 0.5. D) Equal to the product's recycling rate.

Correct: B. A is set in the range 0.2 to 0.8 and is material-specific, reflecting market realities (supply-demand balance for secondary material). Low A prioritises recyclability; high A rewards recycled content.

---

B2 Reference service life under EN 15804 Scenario
Domain: construction | Standards: EN 15804+A2, ISO 15686 | Difficulty: 3 | Verification: TRAINING-MEDIUM

**Scenario.** A facade cladding manufacturer claims a 50-year Reference Service Life (RSL) for their product in an EPD cradle-to-grave declaration. Their warranty is 10 years. Independent field studies show 25-40 year observed service lives depending on climate. The PCR says RSL must be determined per ISO 15686-8 or be consistent with a documented source.

How should the practitioner approach the RSL declaration, and what disclosures are needed? What sensitivity analysis should they perform?

**Rubric (6 points, pass 4):** 1. RSL is not the warranty; warranty is commercial, RSL is technical, these diverge routinely (1 pt). 2. Should apply ISO 15686-8 factor method or use documented literature/standard references, not unsupported claims (1 pt). 3. Must disclose the RSL methodology and assumptions in the EPD (1 pt). 4. Service life impacts modules B (maintenance, repair, refurbishment, replacement) proportionally; longer RSL reduces per-year impact but increases B2-B5 over life (1 pt). 5. Sensitivity analysis on RSL and RSL dependent modules is strongly recommended; show at least two scenarios (1 pt). 6. Flags that climate zone and installation details materially shift service life; a single RSL for global use is a limitation to call out (1 pt).

---

**Scenario.** A facade cladding manufacturer claims a 50-year Reference Service Life (RSL) for their product in an EPD cradle-to-grave declaration. Their warranty is 10 years. Independent field studies show 25-40 year observed service lives depending on climate. The PCR says RSL must be determined per ISO 15686-8 or be consistent with a documented source.

How should the practitioner approach the RSL declaration, and what disclosures are needed? What sensitivity analysis should they perform?

**Rubric (6 points, pass 4):** 1. RSL is not the warranty; warranty is commercial, RSL is technical, these diverge routinely (1 pt). 2. Should apply ISO 15686-8 factor method or use documented literature/standard references, not unsupported claims (1 pt). 3. Must disclose the RSL methodology and assumptions in the EPD (1 pt). 4. Service life impacts modules B (maintenance, repair, refurbishment, replacement) proportionally; longer RSL reduces per-year impact but increases B2-B5 over life (1 pt). 5. Sensitivity analysis on RSL and RSL dependent modules is strongly recommended; show at least two scenarios (1 pt). 6. Flags that climate zone and installation details materially shift service life; a single RSL for global use is a limitation to call out (1 pt).

---

B14 Cross-PCR comparison for ICT equipment Scenario
Domain: electronics | Standards: ISO 14025, ISO 14067 | Difficulty: 2 | Verification: TRAINING-HIGH

**Scenario.** A procurement team wants to compare a server EPD (per EN 50693 PSR for ICT) with a laptop EPD (per PCR from a separate programme operator) on grams CO2e per kWh of compute.

What should the sustainability advisor say?

**Rubric (5 points, pass 4):** 1. ISO 14025 comparability requires same PCR; different PCRs break comparability (1 pt). 2. EN 50693 and the laptop PCR likely differ on system boundary (e.g. embedded manufacturing vs use-phase split), declared vs functional unit, cut-off rules, LCI data vintage, and allocation (1 pt). 3. Even with same PCR, laptop vs server are different products with different compute profiles; g CO2e/kWh of compute requires operational data that EPDs rarely provide (1 pt). 4. Suggests either (a) commissioning a comparative LCA aligned on one PCR, or (b) using the EPDs for single-product reporting only (1 pt). 5. Flags that procurement specifications often require directional comparison; provide informational comparison with strong caveats, not a claim (1 pt).

---

**Scenario.** A procurement team wants to compare a server EPD (per EN 50693 PSR for ICT) with a laptop EPD (per PCR from a separate programme operator) on grams CO2e per kWh of compute.

What should the sustainability advisor say?

**Rubric (5 points, pass 4):** 1. ISO 14025 comparability requires same PCR; different PCRs break comparability (1 pt). 2. EN 50693 and the laptop PCR likely differ on system boundary (e.g. embedded manufacturing vs use-phase split), declared vs functional unit, cut-off rules, LCI data vintage, and allocation (1 pt). 3. Even with same PCR, laptop vs server are different products with different compute profiles; g CO2e/kWh of compute requires operational data that EPDs rarely provide (1 pt). 4. Suggests either (a) commissioning a comparative LCA aligned on one PCR, or (b) using the EPDs for single-product reporting only (1 pt). 5. Flags that procurement specifications often require directional comparison; provide informational comparison with strong caveats, not a claim (1 pt).

---

C9 Temporary carbon storage in long-lived timber products Conflict
Domain: construction | Standards: PAS 2050, ISO 14067, EN 15804+A2 | Difficulty: 3 | Verification: TRAINING-MEDIUM

**Scenario.** A cross-laminated timber (CLT) panel has a claimed 100-year service life in a building. Biogenic CO2 stored in the panel would otherwise return to atmosphere at EoL (incineration/decay). Three standards treat this differently.

Compare PAS 2050, ISO 14067, and EN 15804+A2 on temporary storage. What claims can be defended?

**Rubric (8 points, pass 6):** 1. PAS 2050:2011: historically provided guidance on delayed emissions with a discount factor; has since been revised; check current edition. Temporary storage credit is contested and was narrowed (1 pt). 2. ISO 14067: no default credit for temporary storage; a PCR may allow it if justified and documented; uptake-release tracking is default (1 pt). 3. EN 15804+A2: no temporary storage credit; biogenic CO2 is tracked in GWP-biogenic but in-use storage does not earn credit in Module B; release on EoL is accounted in C3/C4 (1 pt). 4. IPCC AR6 and scientific consensus: temporary storage provides real climate benefit if >decades; but attributional PCFs generally do not credit it (1 pt). 5. Directional effect: a long-life timber product cannot claim "negative lifecycle GWP" under EN 15804+A2 or ISO 14067 from storage alone (1 pt). 6. Communication: marketing claims of "carbon negative" from storage are not supported by the standards and risk greenwash accusations (1 pt). 7. Legitimate option: disclose GWP-biogenic uptake and release, discuss temporary storage in narrative, do not claim a credit (1 pt). 8. False-positive penalty (-1): model must flag that temporary storage is not credited in mainstream PCF standards.

---

**Scenario.** A cross-laminated timber (CLT) panel has a claimed 100-year service life in a building. Biogenic CO2 stored in the panel would otherwise return to atmosphere at EoL (incineration/decay). Three standards treat this differently.

Compare PAS 2050, ISO 14067, and EN 15804+A2 on temporary storage. What claims can be defended?

**Rubric (8 points, pass 6):** 1. PAS 2050:2011: historically provided guidance on delayed emissions with a discount factor; has since been revised; check current edition. Temporary storage credit is contested and was narrowed (1 pt). 2. ISO 14067: no default credit for temporary storage; a PCR may allow it if justified and documented; uptake-release tracking is default (1 pt). 3. EN 15804+A2: no temporary storage credit; biogenic CO2 is tracked in GWP-biogenic but in-use storage does not earn credit in Module B; release on EoL is accounted in C3/C4 (1 pt). 4. IPCC AR6 and scientific consensus: temporary storage provides real climate benefit if >decades; but attributional PCFs generally do not credit it (1 pt). 5. Directional effect: a long-life timber product cannot claim "negative lifecycle GWP" under EN 15804+A2 or ISO 14067 from storage alone (1 pt). 6. Communication: marketing claims of "carbon negative" from storage are not supported by the standards and risk greenwash accusations (1 pt). 7. Legitimate option: disclose GWP-biogenic uptake and release, discuss temporary storage in narrative, do not claim a credit (1 pt). 8. False-positive penalty (-1): model must flag that temporary storage is not credited in mainstream PCF standards.

---

D1 Cement EPD excerpt Spot-the-error
Domain: construction | Standards: EN 15804+A2, EN 16908 | Difficulty: 2 | Verification: TRAINING-MEDIUM

**Excerpt (synthetic).**

> Product: CEM II/A-L 42.5 N Portland-limestone cement. > Declared unit: 1 kg of cement. > Standard: EN 15804+A2. > Scope: Cradle-to-gate (A1-A3) only. > > A1-A3 GWP-fossil: 0.72 kg CO2e > A1-A3 GWP-total: 0.72 kg CO2e > (GWP-biogenic: not applicable) > > This EPD supports comparative claims against other cement EPDs.

**Rubric (8 points, pass 6). Embedded errors:**

1. Declared unit is 1 kg but EN 16908 (PCR for cement) specifies 1 tonne (1000 kg). Wrong declared unit per PCR (1 pt identification, 1 pt justification). 2. GWP sub-indicators incomplete. EN 15804+A2 requires GWP-fossil, GWP-biogenic, GWP-luluc, and GWP-total all to be reported, even if zero. Missing GWP-luluc (1 pt identification, 1 pt justification). 3. "Not applicable" for GWP-biogenic is incorrect phrasing; should be reported as zero with justification (1 pt). 4. Comparative claim from cradle-to-gate A1-A3 only is not valid under ISO 14025 (comparative claims require cradle-to-grave or fully aligned scope and are usually subject to critical review) (1 pt identification, 1 pt justification). 5. False-positive penalty (-1 per wrongly flagged error): no Module D reporting is NOT an error for cradle-to-gate EPDs.

---

**Excerpt (synthetic).**

> Product: CEM II/A-L 42.5 N Portland-limestone cement. > Declared unit: 1 kg of cement. > Standard: EN 15804+A2. > Scope: Cradle-to-gate (A1-A3) only. > > A1-A3 GWP-fossil: 0.72 kg CO2e > A1-A3 GWP-total: 0.72 kg CO2e > (GWP-biogenic: not applicable) > > This EPD supports comparative claims against other cement EPDs.

**Rubric (8 points, pass 6). Embedded errors:**

1. Declared unit is 1 kg but EN 16908 (PCR for cement) specifies 1 tonne (1000 kg). Wrong declared unit per PCR (1 pt identification, 1 pt justification). 2. GWP sub-indicators incomplete. EN 15804+A2 requires GWP-fossil, GWP-biogenic, GWP-luluc, and GWP-total all to be reported, even if zero. Missing GWP-luluc (1 pt identification, 1 pt justification). 3. "Not applicable" for GWP-biogenic is incorrect phrasing; should be reported as zero with justification (1 pt). 4. Comparative claim from cradle-to-gate A1-A3 only is not valid under ISO 14025 (comparative claims require cradle-to-grave or fully aligned scope and are usually subject to critical review) (1 pt identification, 1 pt justification). 5. False-positive penalty (-1 per wrongly flagged error): no Module D reporting is NOT an error for cradle-to-gate EPDs.

---

D4 Battery cell PCF excerpt Spot-the-error
Domain: batteries | Standards: EU Reg. 2023/1542, PEF | Difficulty: 3 | Verification: TRAINING-MEDIUM

**Excerpt (synthetic).**

> Product: NMC811 Li-ion battery cell for EV. > Declared unit: 1 kg of battery cell. > Standard: EU Battery Regulation 2023/1542 and PEF. > Scope: Cradle-to-gate. > > Cathode active material: 8.5 kg CO2e/kg > Anode + separator + electrolyte: 3.2 kg CO2e/kg > Cell assembly + formation: 2.1 kg CO2e/kg > Cradle-to-gate GWP: 13.8 kg CO2e/kg cell > > Recycled content: 25% Co, 15% Ni, 5% Li. CFF A = 0.8 applied. > End-of-life: out of scope (cradle-to-gate declaration).

**Rubric (9 points, pass 6). Embedded errors:**

1. Declared unit "1 kg" is not aligned with EU Battery Regulation PCF methodology. Annex II specifies functional unit as "1 kWh of the total energy provided over the service life of the battery system" (1 pt identification, 1 pt justification). 2. A coefficient 0.8 for battery metals is almost certainly wrong. PEF CFF A for metals is typically 0.2 (reflects saturated scrap market). Using A=0.8 without justification inflates recycled content credit (1 pt identification, 1 pt justification). 3. Cell-level recycled content percentages differ widely by cathode chemistry; applying them without source traceability is inadequate under Battery Regulation (1 pt). 4. Upstream electricity mix for cathode production not disclosed. For NMC, upstream electricity can swing the number by factor 2; required disclosure (1 pt). 5. "Cradle-to-gate declaration" is inconsistent with EU Battery Regulation, which requires cradle-to-gate of the battery "at the factory gate" but with specific inclusion of recycling and manufacturing electricity per Annex II; full cradle-to-gate of battery pack, not just cell (1 pt). 6. No data quality rating disclosed. Battery Regulation requires DQR / data quality class (1 pt). 7. False-positive penalty (-1): "13.8 kg CO2e/kg" is within industry ranges for NMC cells produced on fossil grids; not inherently wrong.

---

**Excerpt (synthetic).**

> Product: NMC811 Li-ion battery cell for EV. > Declared unit: 1 kg of battery cell. > Standard: EU Battery Regulation 2023/1542 and PEF. > Scope: Cradle-to-gate. > > Cathode active material: 8.5 kg CO2e/kg > Anode + separator + electrolyte: 3.2 kg CO2e/kg > Cell assembly + formation: 2.1 kg CO2e/kg > Cradle-to-gate GWP: 13.8 kg CO2e/kg cell > > Recycled content: 25% Co, 15% Ni, 5% Li. CFF A = 0.8 applied. > End-of-life: out of scope (cradle-to-gate declaration).

**Rubric (9 points, pass 6). Embedded errors:**

1. Declared unit "1 kg" is not aligned with EU Battery Regulation PCF methodology. Annex II specifies functional unit as "1 kWh of the total energy provided over the service life of the battery system" (1 pt identification, 1 pt justification). 2. A coefficient 0.8 for battery metals is almost certainly wrong. PEF CFF A for metals is typically 0.2 (reflects saturated scrap market). Using A=0.8 without justification inflates recycled content credit (1 pt identification, 1 pt justification). 3. Cell-level recycled content percentages differ widely by cathode chemistry; applying them without source traceability is inadequate under Battery Regulation (1 pt). 4. Upstream electricity mix for cathode production not disclosed. For NMC, upstream electricity can swing the number by factor 2; required disclosure (1 pt). 5. "Cradle-to-gate declaration" is inconsistent with EU Battery Regulation, which requires cradle-to-gate of the battery "at the factory gate" but with specific inclusion of recycling and manufacturing electricity per Annex II; full cradle-to-gate of battery pack, not just cell (1 pt). 6. No data quality rating disclosed. Battery Regulation requires DQR / data quality class (1 pt). 7. False-positive penalty (-1): "13.8 kg CO2e/kg" is within industry ranges for NMC cells produced on fossil grids; not inherently wrong.

---

E2 Field-level agricultural uncertainty Quantitative OPEN RUBRIC
Domain: food and beverage | Standards: ISO 14067, GHG Protocol Agricultural Guidance | Difficulty: 3 | Verification: TRAINING-MEDIUM

**Scenario.** A UK wheat grower has farm-level data for a PCF. Key inputs:

- Yield: 8.0 t/ha (CoV across 50 farms = 15%). - N fertiliser application: 180 kg N/ha (CoV = 10%). - IPCC Tier 1 direct N2O emission factor: 1.0% of applied N (95% CI: 0.3% to 3.0%, per IPCC 2019 Refinement). - Diesel: 80 L/ha (CoV = 8%).

Propagate uncertainty to cradle-to-gate GWP per tonne wheat. What is the dominant uncertainty driver? How should the result be reported?

**Answer:**

- **Wheat PCF central estimate**: approximately 400-500 kg CO2e/t wheat (order of magnitude). - **Dominant uncertainty driver**: N2O emission factor. The IPCC Tier 1 EF has a 95% CI spanning ~0.3% to 3.0% (factor of 10). N2O has GWP100 of 273 (AR6). Applied N at 180 kg/ha gives direct N2O from 0.5 to 5.4 kg N2O/ha, i.e. 140-1500 kg CO2e/ha. - **Monte Carlo propagation**: lognormal distributions fit EF; normal for yield, fertiliser, diesel. Combined CoV on product PCF likely 30-50%. - **Reporting**: median plus 5th-95th percentile range; clearly state EF uncertainty dominates; recommend Tier 2 (region-specific EF) or Tier 3 (modelled) to reduce uncertainty if material for decision. - **Sensitivity ranking**: EF > yield > fertiliser mass > diesel (typically).

Pass threshold: 4 of 6 points (identify dominant driver, propagate qualitatively, recommend reporting format, recognise EF range magnitude, suggest Tier 2/3, flag GWP metric choice).

---

**Scenario.** A UK wheat grower has farm-level data for a PCF. Key inputs:

- Yield: 8.0 t/ha (CoV across 50 farms = 15%). - N fertiliser application: 180 kg N/ha (CoV = 10%). - IPCC Tier 1 direct N2O emission factor: 1.0% of applied N (95% CI: 0.3% to 3.0%, per IPCC 2019 Refinement). - Diesel: 80 L/ha (CoV = 8%).

Propagate uncertainty to cradle-to-gate GWP per tonne wheat. What is the dominant uncertainty driver? How should the result be reported?

**Answer:**

- **Wheat PCF central estimate**: approximately 400-500 kg CO2e/t wheat (order of magnitude). - **Dominant uncertainty driver**: N2O emission factor. The IPCC Tier 1 EF has a 95% CI spanning ~0.3% to 3.0% (factor of 10). N2O has GWP100 of 273 (AR6). Applied N at 180 kg/ha gives direct N2O from 0.5 to 5.4 kg N2O/ha, i.e. 140-1500 kg CO2e/ha. - **Monte Carlo propagation**: lognormal distributions fit EF; normal for yield, fertiliser, diesel. Combined CoV on product PCF likely 30-50%. - **Reporting**: median plus 5th-95th percentile range; clearly state EF uncertainty dominates; recommend Tier 2 (region-specific EF) or Tier 3 (modelled) to reduce uncertainty if material for decision. - **Sensitivity ranking**: EF > yield > fertiliser mass > diesel (typically).

Pass threshold: 4 of 6 points (identify dominant driver, propagate qualitatively, recommend reporting format, recognise EF range magnitude, suggest Tier 2/3, flag GWP metric choice).

---

C101 Attributional vs consequential LCA — green hydrogen via PEM electrolysis Part C — Conceptual
Domain: energy / hydrogen | Standards: ISO 14040/14044, ISO 14067 | Difficulty: 3 | Verification: TRAINING-HIGH | Source: inspired by UNEP-Q2

**Scenario.** New PEM electrolyser plant in southern Spain, dedicated 200 MW solar farm, intended use: (a) EPD and (b) public "X tonnes CO₂e avoided per kg H₂" claim.

**Rubric (8 points, pass 6):** 1. Defines attributional LCA correctly: bookkeeping of flows physically attributable to the product over its life cycle, average data, status quo (1 pt). 2. Defines consequential LCA correctly: models the change in flows in the wider system caused by the marginal decision to produce one additional kg of H₂; uses marginal/long-run data, market-mediated effects, displaced production (1 pt). 3. Identifies that the **EPD must be attributional** — ISO 14025 / ISO 14067 require attributional bookkeeping; consequential results are not appropriate for product declarations (1 pt). 4. Identifies that the **"avoided emissions vs grey H₂" claim is inherently consequential** — it asks what would happen if the grey H₂ were displaced; this is a marginal decision question, not a bookkeeping one (1 pt). 5. Notes the **electricity attribution problem**: under attributional accounting, the dedicated solar farm electricity is allocated to the H₂ at solar-LCI emissions (~20–40 g CO₂e/kWh embodied PV); under consequential, the right counterfactual is what the solar farm electricity would otherwise have done (likely displaced grid generation), so the claim depends on additionality of the solar build-out (1 pt). 6. Quantifies: attributional kg CO₂e/kg H₂ likely 1–4 kg CO₂e (PV embodied + electrolyser embodied + water + balance-of-plant), versus consequential "avoided" potentially 8–12 kg CO₂e/kg H₂ (vs ~10 kg CO₂e/kg from SMR grey H₂); the two numbers measure different things (1 pt). 7. Flags the **double-counting risk**: claiming both a low attributional EPD figure AND a large avoided-emission credit on the same kg of H₂ inflates the apparent benefit. ISO 14067 §6.4.9.4 / §7 prohibits offsetting against attributional inventories (1 pt). 8. Standards mapping: ISO 14040/14044 §5.2 acknowledges both approaches; ISO 14067 prescribes attributional for product carbon footprints; consequential framing belongs in policy-support documents, not declarations (1 pt).

**False-positive penalty (-1):** if the model claims "consequential is more accurate" or "attributional is outdated" without acknowledging the use-case dependence, score -1.

---

**Scenario.** New PEM electrolyser plant in southern Spain, dedicated 200 MW solar farm, intended use: (a) EPD and (b) public "X tonnes CO₂e avoided per kg H₂" claim.

**Rubric (8 points, pass 6):** 1. Defines attributional LCA correctly: bookkeeping of flows physically attributable to the product over its life cycle, average data, status quo (1 pt). 2. Defines consequential LCA correctly: models the change in flows in the wider system caused by the marginal decision to produce one additional kg of H₂; uses marginal/long-run data, market-mediated effects, displaced production (1 pt). 3. Identifies that the **EPD must be attributional** — ISO 14025 / ISO 14067 require attributional bookkeeping; consequential results are not appropriate for product declarations (1 pt). 4. Identifies that the **"avoided emissions vs grey H₂" claim is inherently consequential** — it asks what would happen if the grey H₂ were displaced; this is a marginal decision question, not a bookkeeping one (1 pt). 5. Notes the **electricity attribution problem**: under attributional accounting, the dedicated solar farm electricity is allocated to the H₂ at solar-LCI emissions (~20–40 g CO₂e/kWh embodied PV); under consequential, the right counterfactual is what the solar farm electricity would otherwise have done (likely displaced grid generation), so the claim depends on additionality of the solar build-out (1 pt). 6. Quantifies: attributional kg CO₂e/kg H₂ likely 1–4 kg CO₂e (PV embodied + electrolyser embodied + water + balance-of-plant), versus consequential "avoided" potentially 8–12 kg CO₂e/kg H₂ (vs ~10 kg CO₂e/kg from SMR grey H₂); the two numbers measure different things (1 pt). 7. Flags the **double-counting risk**: claiming both a low attributional EPD figure AND a large avoided-emission credit on the same kg of H₂ inflates the apparent benefit. ISO 14067 §6.4.9.4 / §7 prohibits offsetting against attributional inventories (1 pt). 8. Standards mapping: ISO 14040/14044 §5.2 acknowledges both approaches; ISO 14067 prescribes attributional for product carbon footprints; consequential framing belongs in policy-support documents, not declarations (1 pt).

**False-positive penalty (-1):** if the model claims "consequential is more accurate" or "attributional is outdated" without acknowledging the use-case dependence, score -1.

---

C107 Laptop cradle-to-gate carbon footprint — answer in **kg CO₂e** Part C — Quantitative
Domain: electronics | Standards: ISO 14067, manufacturer EPDs (Apple, Dell, Lenovo PCFs) | Difficulty: 2 | Verification: TRAINING-MEDIUM | Source: inspired by UNEP-Q13

**Scenario.** High-end 16-inch laptop with discrete GPU, 2024–2026 generation. Answer in **kg CO₂e**.

**Answer key:** - **Defensible cradle-to-gate range: ~250 to ~450 kg CO₂e per unit.** Single-number answer in the 300–400 kg CO₂e range is acceptable. - Manufacturer-published cradle-to-gate cell-and-board-dominated estimates (Apple MacBook Pro 16" Environmental Reports 2023–2024; Dell XPS 16; Lenovo ThinkPad P-series PCFs) cluster in this band. - **Dominant contributors** (must be named): integrated circuits (CPU + GPU + memory), display assembly (especially OLED / mini-LED), aluminium / magnesium chassis, battery cell manufacturing. - **Citations** must be real (manufacturer Environmental Report or peer-reviewed paper). Acceptable: Apple Environmental Report 2024, Dell PCF disclosures, Teehan & Kandlikar (2013) for older baseline, IEA/IRENA electronics LCAs, Suckling & Lee (2015) for component-level breakdown. - **Uncertainty range** must be stated: at least ±30% reasonable; full publishable range is wider (e.g. 200–600 kg CO₂e/unit) once spec variability and grid-electricity assumptions are included.

**Rubric (6 points, pass 4):** 1. Numeric answer in **300–400 kg CO₂e** (full credit) or 200–500 kg CO₂e (partial, 0.5 pt) (1 pt). 2. Unit stated correctly as **kg CO₂e** matching the question (1 pt). 3. Identifies ICs / display / battery as dominant contributors (1 pt). 4. Cites at least one real, traceable source (manufacturer report or peer-reviewed paper) (1 pt). 5. States an uncertainty range (1 pt). 6. Acknowledges that a "single-laptop" PCF depends on the spec and is meaningless without a functional-unit definition (per device vs per device-year of use) (1 pt).

**False-positive penalty (-2):** answer in **tonnes** while the question asked for kg → score -2 (the model has fallen into the unit trap).

**Cross-question linkage:** see C108. The two questions test consistency — a model that answers ~350 kg CO₂e here and then ~350 tonnes (or any number with "tonnes" unit) on C108 has confused units catastrophically.

---

**Scenario.** High-end 16-inch laptop with discrete GPU, 2024–2026 generation. Answer in **kg CO₂e**.

**Answer key:** - **Defensible cradle-to-gate range: ~250 to ~450 kg CO₂e per unit.** Single-number answer in the 300–400 kg CO₂e range is acceptable. - Manufacturer-published cradle-to-gate cell-and-board-dominated estimates (Apple MacBook Pro 16" Environmental Reports 2023–2024; Dell XPS 16; Lenovo ThinkPad P-series PCFs) cluster in this band. - **Dominant contributors** (must be named): integrated circuits (CPU + GPU + memory), display assembly (especially OLED / mini-LED), aluminium / magnesium chassis, battery cell manufacturing. - **Citations** must be real (manufacturer Environmental Report or peer-reviewed paper). Acceptable: Apple Environmental Report 2024, Dell PCF disclosures, Teehan & Kandlikar (2013) for older baseline, IEA/IRENA electronics LCAs, Suckling & Lee (2015) for component-level breakdown. - **Uncertainty range** must be stated: at least ±30% reasonable; full publishable range is wider (e.g. 200–600 kg CO₂e/unit) once spec variability and grid-electricity assumptions are included.

**Rubric (6 points, pass 4):** 1. Numeric answer in **300–400 kg CO₂e** (full credit) or 200–500 kg CO₂e (partial, 0.5 pt) (1 pt). 2. Unit stated correctly as **kg CO₂e** matching the question (1 pt). 3. Identifies ICs / display / battery as dominant contributors (1 pt). 4. Cites at least one real, traceable source (manufacturer report or peer-reviewed paper) (1 pt). 5. States an uncertainty range (1 pt). 6. Acknowledges that a "single-laptop" PCF depends on the spec and is meaningless without a functional-unit definition (per device vs per device-year of use) (1 pt).

**False-positive penalty (-2):** answer in **tonnes** while the question asked for kg → score -2 (the model has fallen into the unit trap).

**Cross-question linkage:** see C108. The two questions test consistency — a model that answers ~350 kg CO₂e here and then ~350 tonnes (or any number with "tonnes" unit) on C108 has confused units catastrophically.

---

The full benchmark

The full report covers all 111 questions, the complete rubric, per-model verdicts on the canonical cohort, the seven-criterion scoring legend, the two-judge agreement analysis, and the methodology paper text. Delivered as PDF within 60 seconds. CC-BY 4.0.

Full report · PDF

Get the full report

One email, one PDF. We use your email to send the report and, if you opt in, occasional updates when new models are graded or rubric items resolve. Unsubscribe with one click.

Sent from noreply@verdatir.com. We store your address to deliver the report and, if you opt in, future updates. See /privacy. CC-BY 4.0.
Welcome back. You've already requested the full report.

Download the report (PDF) ↓