Earnings Call Summarization Cost: Examples
Covering a 50-ticker earnings universe costs single-digit dollars annually on budget models and under twenty dollars on premium ones. That is the surprising finding these scenarios document. The workload is fixed: 50 tickers, one 20,000-token transcript each, 1,500-token output, half the input cached, four quarters a year. Model choice drives the cost; the ceiling is lower than most practitioners expect.
Worked Examples
See the inputs and outcome together
Each scenario keeps the starting point, the outcome, and the actual lesson in one place so the page reads like a decision notebook, not a data dump.
- 1
Cheapest tier across the whole universe
Summarizing all 50 tickers every quarter on Gemini 2.5 Flash-Lite. This is the floor for a bulk, first-pass summarization job.
Cost per stock per quarter $0.00185, full-year total across the universe $0.37.
Model
Gemini 2.5 Flash-Lite
Tickers per quarter
50
Transcript tokens
20,000
Summary tokens
1,500
Cache hit rate
50%
Thirty-seven cents covers 200 transcript summaries for a full year. At this price the bottleneck is never the model bill; it is transcript ingestion and quality control. Cost should not drive you off the cheap tier for a coarse summary.
- 2
Mid-tier quality jump
Stepping up to Claude Haiku 4.5 for cleaner summaries, same universe and same caching. The first point where quality, not price, is the reason to move.
Cost per stock per quarter $0.0185, full-year total $3.70.
Model
Claude Haiku 4.5
Tickers per quarter
50
Transcript tokens
20,000
Summary tokens
1,500
Cache hit rate
50%
Haiku is ten times Flash-Lite, but ten times $0.37 is still only $3.70 for the year. The absolute numbers are so small that you should choose on summary quality and downstream error rate, not on the model bill.
- 3
Flagship model, full universe
The top of the table: Claude Opus 4.7 summarizing all 50 tickers every quarter. The most you would plausibly pay for the highest-fidelity summaries.
Cost per stock per quarter $0.0925, full-year total $18.50.
Model
Claude Opus 4.7
Tickers per quarter
50
Transcript tokens
20,000
Summary tokens
1,500
Cache hit rate
50%
Even the flagship costs only $18.50 a year for the entire universe, fifty times the cheapest tier yet still trivial. Summarization is dominated by short output, so the output-price gap between models barely matters at this scale.
Patterns
Try These Tools
Run the numbers next
Financial Document Token Estimator
Paste a 10-K, 10-Q, 8-K or earnings transcript and see token count + one-pass extraction cost across eight frontier LLMs, with cache-hit toggle.
Token-Cost Optimizer
Compute the dollar cost of a trading research loop across Claude, GPT, and Gemini. Prompt length × model × retry × call volume → cost per idea and per.
Sources & References
- Anthropic API Pricing — Anthropic (2026)
- Prompt Caching with Claude — Anthropic (2026)
Related Content
Keep the topic connected
Agent-Cost Envelope
The agent-cost envelope: the loop of (calls × tokens × retries × model_price) that determines the dollar cost of an LLM-driven trading agent per decision.
Hallucination Detection
Detecting LLM hallucinations in financial outputs: the verifiable-claim approach, citation grounding, and cross-model agreement signals that work.
LLM for Finance Deployment Checklist
A pre-flight checklist for putting a large language model into a finance workflow: scoping, grounding, input security, numerical verification, and drift monitoring.