Head-to-head showdown: Cohere Command A+ ($2.50 in / $10.00 out per 1M) vs DeepInfra — Llama 3.3 70B ($0.10 in / $0.30 out per 1M). DeepInfra — Llama 3.3 70B is 31.3× cheaper across standard token mixes, with 128K vs 128K context windows.
command-a-plus · cohere
meta-llama/Llama-3.3-70B-Instruct · deepinfra
Benchmark
| Workload Scenario | Cohere Command A+ | DeepInfra — Llama 3.3 70B | Price Delta |
|---|---|---|---|
| 1M input tokens (raw text) | $2.50 | $0.10 | +$2.40 |
| 1M output tokens (generation) | $10.00 | $0.30 | +$9.70 |
| 1M tokens · 70% input / 30% output mix | $4.75 | $0.16 | +$4.59 |
| Standard chat turn (4K in / 800 out, 50% cached) | $0.0135 | $0.00064 | +$0.0129 |
| Monthly scale (10K requests / day) | $4,050.00 | $192.00 | +$3,858.00 |
Negative difference = Cohere Command A+ is cheaper. Positive = DeepInfra — Llama 3.3 70B is cheaper.
Scaling Curve
Total cost of a token volume at 70/30 input/output split (uncached). Log-log scale.
Verdict
DeepInfra — Llama 3.3 70B is cheaper on both input and output rates, so it costs less at every input/output mix. Price alone still isn't the whole decision: capability, latency and context limits (Cohere Command A+: 128K, DeepInfra — Llama 3.3 70B: 128K) may justify the premium for your task.
FAQ
On input, DeepInfra — Llama 3.3 70B is cheaper ($0.10/M vs $2.50/M). On output, DeepInfra — Llama 3.3 70B is cheaper ($0.30/M vs $10.00/M). The same model is cheaper on both sides, so it wins at every mix.
A chat-style request (4,000 input + 800 output tokens, 50% cached) costs $0.0135 on Cohere Command A+ and $0.00064 on DeepInfra — Llama 3.3 70B — DeepInfra — Llama 3.3 70B is 21.1× more expensive for that workload.
Cohere Command A+ supports 128,000 tokens (8.2K max output); DeepInfra — Llama 3.3 70B supports 128,000 (8.2K max output). They are equal.