HAIL.SO / DISPATCH · 2026-07-12 · COMPARE
FILE: COMPARE · 6 of 6 slots

side by side.

← all costsclear
01

Large language models

3 models
Mistral
Codestral 25.08
×
Perplexity
Sonar
×
DeepSeek
DeepSeek V3
×
Output $/MTok$0.90$1.00$0.28
Input $/MTok$0.30$1.00$0.14
Cached input $/MTok$0.0028
Context window131,072128,0001,048,576
Output cap8,1928,000384,000
Modalitiesin: text / out: textin: text / out: textin: text / out: text
Tool use
Structured output
FamilyCodestralSonarDeepSeek V3
Knowledge cutoff
Verified2026-07-022026-07-022026-07-02
Sourceprovider page ↗provider page ↗provider page ↗
02

Speech to text

1 model
Google
Chirp 2
×
$/min$0.016000
$/min batch
Streaming
Realtime
Languages20+
Diarizationincluded
TTFW
Verified2026-07-02
Sourceprovider page ↗
03

Text to speech

2 models
Smallest.ai
Lightning v3.1
×
Deepgram
Deepgram Aura 2
×
$/1M chars$17.50$30.00
Voice qualityneuralneural
Voice cloning✓ included
Voice count217
Languagesen, hi, es, mr, kn, ta, bn, gu, te, ml, pa, oren, es, de, fr, nl, it, ja
SSML support
TTFB200ms
Output formatspcm, mp3, wav, ulaw, alawwav, mp3, linear16, mulaw, alaw, opus
Verified2026-07-022026-07-02
Sourceprovider page ↗provider page ↗