Baidu: ERNIE-Light-DT-1.0 vs Deepseek V4 Pro — API Pricing, 150K Cache & Latency Benchmark
Side-by-side unit economics audit comparing Baidu: ERNIE-Light-DT-1.0 (baidu/ernie-light-dt-1.0-intl: Input $1.13/1M, Output $3.94/1M, Warm Cache Read $0.113/1M) against Deepseek V4 Pro (deepseek-v4-pro: Input $0.66/1M, Output $1.98/1M, Warm Cache Read $0.066/1M).
Activate 4-breakpoint prefix caching on Baidu: ERNIE-Light-DT-1.0 to cut input token costs from $1.13/1M down to $0.113/1M, or route async workloads via Batch API for 50% off.
Lock static system prompt prefixes on Deepseek V4 Pro to pay $0.066/1M warm cache rates instead of $0.66/1M uncached input, plus eligible startup cloud credits.
In Etuki FinOps telemetry comparing Baidu: ERNIE-Light-DT-1.0 (API ID: baidu/ernie-light-dt-1.0-intl) vs Deepseek V4 Pro (API ID: deepseek-v4-pro), Deepseek V4 Pro offers lower standard uncached input pricing at $0.66/1M tokens, while Deepseek V4 Pro leads in 10-turn 150K cache-pinned coding economics at $0.2524 per 10 turns (saving $0.1484 over Baidu: ERNIE-Light-DT-1.0). For latency-sensitive interactive agents, Deepseek V4 Pro delivers 42ms faster P90 time-to-first-token (390ms vs 432ms).
| Specification / FinOps Dimension | Baidu: ERNIE-Light-DT-1.0 (baidu/ernie-light-dt-1.0-intl) | Deepseek V4 Pro (deepseek-v4-pro) | Verified Advantage |
|---|---|---|---|
| Official Vendor API ID | baidu/ernie-light-dt-1.0-intl | deepseek-v4-pro | UMCE 563-Registry Locked |
| 563-Registry & Routing Eligibility | is_known: true | routable: true | is_known: true | routable: true | Both Merkle Verified |
| Standard Uncached Input (/1M tokens) | $1.13 / 1M | $0.66 / 1M | Deepseek V4 Pro |
| Warm Prefix Cache Read (/1M tokens) | $0.113 / 1M | $0.066 / 1M | Deepseek V4 Pro |
| Standard Output (/1M tokens) | $3.94 / 1M | $1.98 / 1M | Deepseek V4 Pro |
| Context Window Capacity | 1024K Tokens | 200K Tokens | Baidu: ERNIE-Light-DT-1.0 |
| 150K Prompt × 10 Turns (Unpinned Cold + 2K Out) | $1.7738 | $1.0296 | Deepseek V4 Pro |
| 150K × 10 Turns Case A (100% Static Pinned) | $0.4008 (saves $1.3730) | $0.2524 (saves $0.7772) | Deepseek V4 Pro (-$0.1484) |
| 150K × 10 Turns Case B (90% Static + 10% Tail) | $0.5381 (saves $1.2357) | $0.3326 (saves $0.6970) | Deepseek V4 Pro |
| P50 / P90 First-Token Latency | 294ms / 432ms (89 tok/s) | 265ms / 390ms (96 tok/s) | Deepseek V4 Pro (-42ms) |
1. Full-Turn Rate-Card Substitution for Baidu: ERNIE-Light-DT-1.0: At an uncached input rate of $1.13/1M, output rate of $3.94/1M, and warm cache read rate of $0.113/1M, running 10 consecutive 150,000-token coding agent turns (with 2,000 output tokens/turn) without prefix pinning costs $1.7738. Locking 100% static prefixes (Case A: 1 cold turn + 9 warm cached turns + 20K total output) reduces 10-turn spend to $0.4008 (saving $1.3730); under a 90% static prefix + 10% dynamic tail workload (Case B), 10-turn spend is $0.5381.
2. Full-Turn Rate-Card Substitution for Deepseek V4 Pro & Winner Verdict: On Deepseek V4 Pro ($0.66/1M input, $1.98/1M output, $0.066/1M warm cache), 10 unpinned 150K turns cost $1.0296, dropping to $0.2524 in Case A and $0.3326 in Case B. Across every 10-turn 150K coding loop (including output tokens), Deepseek V4 Pro delivers an additional $0.1484 Case-A cost advantage.
Production Workload TCO: 25,000 MAU • 50M Tokens/Mo (50% Cache Hit) • 2,000 GB Egress
💡 Optimal FinOps Verdict: Deploying Deepseek V4 Pro saves $29.94/month ($359/year) at 25,000 MAU & 50M monthly tokens (80% input / 20% output split with 50% prefix cache hit rate).