bargo/transcribe-finance-v1 is Bargo's proprietary English speech‑to‑text model built for corporate earnings calls, institutional investor conferences, FOMC commentary, financial podcasts, and biopharma discussions. It delivers 2.38% WER on SPGISpeech and 3.92% composite WER across OpenASR at $0.10 per audio hour.
01 · Standardized Evaluation
Evaluated on standard public speech-to-text suites covering U.S. corporate earnings calls (SPGISpeech), clean and noisy/accented speech (LibriSpeech), and global multi-accent earnings webcasts (Earnings-22). Lower Word Error Rate (WER %) is better.
| Benchmark Suite | Bargo v1 ($0.10/h) | Bargo Realtime ($0.05/h) | Deepgram Nova-3 | AssemblyAI U-2 | MAI-Transcribe-1 | GPT-4o Transcribe | Whisper Large-v3 |
|---|---|---|---|---|---|---|---|
| SPGISpeechCorporate Earnings Calls | 2.38% | 5.43% | 2.95% | 3.10% | 3.20% | 3.50% | 3.92% |
| LibriSpeech CleanClean English Speech | 1.48% | 4.22% | 1.83% | 1.60% | 1.75% | 1.90% | 1.84% |
| LibriSpeech OtherNoisy & Accented Speech | 2.66% | 7.44% | 3.90% | 3.10% | 3.40% | 3.60% | 3.68% |
| Earnings-22Global Multi-Accent Earnings | 9.15% | 17.44% | 7.84% | 8.90% | 9.00% | 9.60% | 11.30% |
| 4-Suite Mean WERComposite Average | 3.92% | 8.63% | 4.13% | 4.18% | 4.34% | 4.65% | 5.19% |
| Spoken Stock TickersCase-Exact Exchange Symbols | 100.0% (0.43% WER) | 87.5% (4.76% WER) | 67.5% | 70.0% | 72.5% | 75.0% | 65.0% |
| 16-Call / 201-Exec RosterC-Suite & Sell-Side Recall | 99.0% (199/201) | 88.4% (8.18% WER) | 68%–74% | 65%–72% | — | 70%–76% | 62.0% |
| PriMock57 Clinical RecallBiopharma & Clinical Terms | 98.2% (8.83% WER) | 85.5% (12.09% WER) | — | — | — | — | — |
| Retail Price / HrEffective Cost per Audio Hour | $0.10 / hr | $0.05 / hr | $0.26 / hr | $0.15 / hr | $0.10 / hr | $0.36 / hr | $0.36 / hr |
02 · Financial & Biopharma Fidelity
Off-the-shelf ASR models routinely corrupt executive names, spoken stock tickers, basis-point figures, and clinical trial assets. bargo/transcribe-finance-v1 resolves domain terminology natively and accepts optional prompt hints for call rosters.
Exact spelling of CEOs, CFOs, IR heads, and covering sell-side analysts during live earnings Q&A.
Resolves spoken exchange symbols (NVDA, LLY, ASML, TSM, BRK.A) without phonetic distortion.
Canonical formatting for GAAP/non-GAAP, EBITDA, basis points, EPS ranges, and YoY/QoQ metrics.
Validated on PriMock57 clinical consultations, therapeutic classes, and FDA AdCom terminology.
03 · Drop-In Integration
100% compatible with the standard OpenAI Audio Transcriptions SDK (/v1/audio/transcriptions) and Multimodal Chat Completions (/v1/chat/completions).
from openai import OpenAI client = OpenAI( base_url="https://www.bargo.ai/v1", api_key="YOUR_BARGO_API_KEY", ) with open("earnings_call.mp3", "rb") as audio_file: transcript = client.audio.transcriptions.create( model="bargo/transcribe-finance-v1", file=audio_file, prompt="Eli Lilly, Anat Ashkenazi, David Ricks, Zepbound, Mounjaro", ) print(transcript.text)
curl -X POST https://www.bargo.ai/v1/audio/transcriptions \ -H "Authorization: Bearer $BARGO_API_KEY" \ -F file="@earnings_call.mp3" \ -F model="bargo/transcribe-finance-v1" \ -F prompt="Eli Lilly, Anat Ashkenazi, Zepbound, Mounjaro" \ -F response_format="verbose_json"
curl -X POST https://www.bargo.ai/v1/chat/completions \ -H "Authorization: Bearer $BARGO_API_KEY" \ -H "Content-Type: application/json" \ -d '{ "model": "bargo/transcribe-finance-v1", "messages": [ { "role": "user", "content": [ {"type": "text", "text": "Eli Lilly, Anat Ashkenazi, Zepbound, Mounjaro"}, {"type": "input_audio", "input_audio": {"data": "<BASE64_AUDIO>", "format": "mp3"}} ] } ] }'
import fs from "fs"; import OpenAI from "openai"; const client = new OpenAI({ baseURL: "https://www.bargo.ai/v1", apiKey: process.env.BARGO_API_KEY, }); const result = await client.audio.transcriptions.create({ model: "bargo/transcribe-finance-v1", file: fs.createReadStream("earnings_call.mp3"), prompt: "NVIDIA, Jensen Huang, Colette Kress, Blackwell, NVLink", }); console.log(result.text);
04 · Model Datasheet
Production parameters and compliance guarantees for institutional deployments.
bargo/transcribe-finance-v1 into your research or voice pipeline?