This page uses a curated launch benchmark slice from the broader Tokelang Lite evaluation corpus: strong compaction examples across incident response, legal, sales, analytics, research, and operations, plus two deliberate passthrough cases where exact wording should stay intact.
POST /v1/compile with prompt only. Token counts use tiktoken cl100k_base, and latency uses Client-measured end-to-end HTTPS latency to the live public deployment from the production droplet.
Passthrough is not a miss. It is the compiler choosing fidelity over compression for code, exact math, row-heavy inputs, and contract-sensitive prompts.
Each prompt was measured with 1 warmup + 5 measured requests. The table below shows average latency and p95 latency per prompt.
These examples are selected to show both sides of the product: compaction on stronger workflow-style prompts, and passthrough when exact wording should not be compressed.
34 to 15 tokens, 55.88% savings, 11.86 ms average latency.
Turn the sales call notes into follow-up actions. Extract buyer pain points, timeline, budget signals, procurement blockers, and return a short account strategy memo for the account executive.
process analyze buyer pain points timeline budget signals procurement blockers account strategy memo executive brief
48 to 26 tokens, 45.83% savings, 11.69 ms average latency.
Contract review workflow. Step 1: Read the payment terms. Step 2: If the side letter changes payment timing, surface the exception. Step 3: Otherwise compare warranty and indemnity language. Step 4: Return a short legal memo for procurement.
process default simple analyze payment terms analyze side letter changes payment timing surface exception else compare warranty indemnity language output return legal memo for procurement brief
25 to 25 tokens, 0.00% savings, 10.00 ms average latency.
Translate this customer message into German. Keep the tone polite, preserve the order number exactly, and return only the translated text.
Translate this customer message into German. Keep the tone polite, preserve the order number exactly, and return only the translated text.
37 to 37 tokens, 0.00% savings, 10.28 ms average latency.
Solve this exactly. Let f(x) = x^4 - 6x^2 + 8x - 3. Find all critical points, classify them, and keep the symbolic derivation explicit.
Solve this exactly. Let f(x) = x^4 - 6x^2 + 8x - 3. Find all critical points, classify them, and keep the symbolic derivation explicit.
This public table keeps the cleaner launch-facing examples and two deliberate passthrough cases. Expected mode reflects the benchmark design intent. Actual mode reflects what the current benchmarked build returned during measurement.
| Prompt | Expected | Actual | Original | Output | Savings | Avg latency | p95 latency |
|---|---|---|---|---|---|---|---|
Sales Follow-Up Strategy pb-06 · sales | tokelang | tokelang | 34 | 15 | 55.88% | 11.86 ms | 12.76 ms |
Security Alert Triage pb-14 · security | tokelang | tokelang | 40 | 18 | 55.00% | 11.36 ms | 12.10 ms |
User Research Themes pb-15 · user-research | tokelang | tokelang | 34 | 16 | 52.94% | 10.81 ms | 11.59 ms |
Contract Exception Memo pb-08 · legal | tokelang | tokelang | 48 | 26 | 45.83% | 11.69 ms | 12.05 ms |
Grant Proposal Review pb-16 · grants | tokelang | tokelang | 27 | 16 | 40.74% | 18.60 ms | 39.94 ms |
Exec Meeting Summary pb-04 · operations | tokelang | tokelang | 29 | 18 | 37.93% | 11.20 ms | 11.90 ms |
Incident Response Narrative pb-02 · incident-response | tokelang | tokelang | 52 | 33 | 36.54% | 10.24 ms | 10.79 ms |
Lesson Plan pb-10 · education | tokelang | tokelang | 33 | 23 | 30.30% | 10.26 ms | 10.48 ms |
Exact Translation pb-17 · translation | passthrough | passthrough | 25 | 25 | 0.00% | 10.00 ms | 10.50 ms |
Exact Symbolic Math pb-21 · math | passthrough | passthrough | 37 | 37 | 0.00% | 10.28 ms | 11.46 ms |