Claude Opus vs Llama 4 Maverick: Summarize a 3,000-Word Article
The Prompt
Summarize the following 3,000-word article about the global semiconductor supply chain crisis into a concise executive briefing. Include: (1) the 3 most critical bottlenecks, (2) which companies are most affected, (3) projected timeline for recovery, and (4) recommended actions for mid-size tech companies. Keep it under 400 words.
This battle tests real-world summarization skills — the kind of task millions of professionals run daily. We fed both models the same long-form article about semiconductor supply chains (a complex, data-heavy topic) and asked for a structured executive briefing. Good summarization means capturing the right details, maintaining accuracy, and delivering a scannable format — not just shortening text.
Claude Opus 4.6Llama 4 Maverick
## Executive Briefing: Semiconductor Supply Chain Crisis (April 2026)\n\n### 3 Critical Bottlenecks\n\n1. **Advanced packaging capacity** — TSMC and Samsung control 78% of CoWoS (Chip-on-Wafer-on-Substrate) production. Demand from AI accelerators has created an 18-month backlog. New facilities in Arizona and Kumamoto won't reach volume production until Q3 2027.\n\n2. **Neon gas supply concentration** — Despite diversification efforts post-2022, 43% of semiconductor-grade neon still flows through three suppliers. Any regional disruption creates cascading shortages within 6-8 weeks.\n\n3. **EUV lithography bottleneck** — ASML remains the sole EUV supplier. Their 2026 delivery schedule is fully booked, with High-NA EUV systems allocated exclusively to Intel, TSMC, and Samsung through 2028.\n\n### Most Affected Companies\n\n- **Nvidia & AMD**: AI chip demand outstrips packaging capacity by 2.3x. Both face 6-9 month lead times for data center GPUs.\n- **Automotive OEMs (Toyota, VW, Stellantis)**: Mature-node chips (28nm+) still face spot shortages. Vehicle production remains 8-12% below 2019 levels.\n- **Apple & Qualcomm**: Advanced mobile SoCs competing for the same TSMC 2nm capacity, creating allocation friction.\n\n### Recovery Timeline\n\n- **Mature nodes (28nm+)**: Largely resolved by Q4 2026 as Chinese fabs (SMIC, Hua Hong) add capacity.\n- **Advanced nodes (5nm and below)**: Tight through 2027. Relief begins Q3 2027 with TSMC Arizona Phase 1 and Samsung Taylor ramp.\n- **Advanced packaging**: Most constrained segment. Full demand-supply equilibrium unlikely before 2028.\n\n### Recommended Actions for Mid-Size Tech Companies\n\n1. **Dual-source critical components** — Qualify at least two foundries for each chip. Budget 15-20% premium for supply security.\n2. **Extend planning horizons** — Move from 6-month to 18-month procurement cycles for any advanced-node components.\n3. **Evaluate chiplet architectures** — Chiplet designs reduce dependency on cutting-edge nodes and improve packaging flexibility.\n4. **Build strategic inventory** — Maintain 12-16 weeks of safety stock for sole-source components (up from the typical 4-6 weeks).\n\n*Word count: 312*
Llama 4 MaverickLlama 4 Maverick
## Semiconductor Supply Chain: Executive Summary\n\n**Key Bottlenecks:**\n\nThe article identifies three main bottlenecks in the current semiconductor supply chain:\n\n1. **Packaging and Testing** - There is a significant shortage of advanced packaging capacity, particularly for AI chips. TSMC's CoWoS technology is in extremely high demand and production cannot keep pace. This affects the entire downstream supply chain.\n\n2. **Raw Material Dependencies** - Critical materials like neon gas and certain rare earth elements remain concentrated in a few regions. Supply chain diversification has been slow despite years of effort.\n\n3. **Equipment Availability** - ASML's lithography machines, especially EUV systems, have long lead times and limited production capacity. This creates a fundamental constraint on how fast the industry can expand.\n\n**Affected Companies:**\n\nThe biggest impact falls on AI chip makers like Nvidia and AMD who face extended lead times. Automotive companies continue to deal with mature node shortages, though the situation is improving. Apple and other smartphone makers face competition for advanced node allocation at leading foundries.\n\n**Recovery Timeline:**\n\nFor older chip technologies (28nm and above), supply should normalize by late 2026 or early 2027. Advanced nodes will remain tight through 2027 as new fabs come online in the US and Japan. Advanced packaging is the longest pole in the tent and may not reach balance until 2028.\n\n**Recommendations for Mid-Size Companies:**\n\n- Diversify suppliers where possible\n- Plan further ahead for chip procurement\n- Consider alternative chip architectures that don't require cutting-edge manufacturing\n- Keep more inventory on hand as a buffer\n\nThe overall outlook is cautiously optimistic, with significant new capacity coming online, but companies should not expect a return to pre-shortage conditions in the near term.\n\n*Word count: 267*
🔍 Analysis
Both models delivered functional summaries, but the quality gap is significant.\n\nStructure & Specificity: Claude Opus nailed the structured executive briefing format with precise data points — "78% of CoWoS production," "18-month backlog," "2.3x demand-to-capacity ratio," "8-12% below 2019 levels." Llama 4 Maverick produced a competent summary but leaned heavily on generalities like "extremely high demand" and "long lead times" without quantifying them.\n\nAccuracy & Detail: Claude's recovery timeline included specific dates (Q3 2027, Q4 2026) tied to specific events (TSMC Arizona Phase 1, Samsung Taylor ramp). Llama's timeline was vaguer — "late 2026 or early 2027" without anchoring to specific capacity additions.\n\nActionability: The recommendations section is where Claude pulled furthest ahead. Specific numbers (15-20% premium, 12-16 weeks safety stock, 18-month procurement cycles) give a reader something to act on immediately. Llama's recommendations were correct but generic — "diversify suppliers" and "keep more inventory" don't tell a CFO what to budget.\n\nReadability: Both were well-organized and scannable. Llama's prose was slightly more conversational, which some readers may prefer. Claude's was denser but more information-rich per sentence.\n\nWhere Llama held up: The summary was accurate, well-structured, and hit all four requested sections. At roughly 1/30th the API cost, Llama 4 Maverick delivers solid summarization for high-volume, cost-sensitive workflows. The closing editorial line ("cautiously optimistic") was a nice touch.\n\nFor executive-level deliverables where precision matters, Claude Opus is the clear winner. For bulk summarization at scale where "good enough" saves thousands in API costs, Llama 4 Maverick is a legitimate choice.
"We tested Claude Opus vs Llama 4 Maverick on summarizing a 3,000-word article. Claude's summary had 4x more specific data points. But Llama costs 30x less. Which matters more?"