RedRouterAbliterated Model API Speed, Reliability and Price
Every 8 Hours from US East & West72-Hour WindowUpdated
Provider comparison
DeepSeek V4.1 Flash 1017- DeepSeek V4.1 Flash (undated derivative)
GLM 5.3 Flash- GLM 5.3
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
 Experiential Labsdealignai |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.394 | $0.225 | $0.900 | 223.8 | 0.6 | 2.9 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:08 UTC |
  Heabsy via NanoGPT |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.610 | $0.350 | $1.390 | — | — | — | 0.0%0 / 72 requests | failing | 08 Nov 2026, 10:13 UTC |
 OrcaRouter via NanoGPT |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $1.050 | $0.840 | $1.680 | 154.3 | 1.5 | 4.7 | 22.2%16 / 72 requests | failing | 09 Nov 2026, 02:13 UTC |
 Sable via NanoGPT |
DeepSeek V4.1 Flash 1017 | heretic | NATIVE | 512,000 | $0.303 | $0.173 | $0.693 | 110.7 | 3.8 | 8.7 | 95.8%69 / 72 requests | ok | 11 Nov 2026, 10:15 UTC |
 OpenRoutermorph/fp8 |
DeepSeek V4.1 Flash 1017 | none | FP8 | 1,048,576 | $0.117 | $0.022 | $0.400 | 91.4 | 0.7 | 6.9 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:13 UTC |
 Venice |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.578 | $0.330 | $1.320 | 183.8 | 0.5 | 3.5 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:07 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- Experiential Labs · dealignai
- Heabsy via NanoGPT
- OrcaRouter via NanoGPT
- Sable via NanoGPT
- OpenRouter · morph/fp8
- Venice
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| Experiential Labsdealignai |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.394 | $0.225 | $0.900 | 224.0 | 0.6 | 2.9 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:08 UTC |
| Heabsy via NanoGPT |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.610 | $0.350 | $1.390 | — | — | — | 0.0%0 / 36 requests | failing | 08 Nov 2026, 10:13 UTC |
| OrcaRouter via NanoGPT |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $1.050 | $0.840 | $1.680 | 154.3 | 1.5 | 4.5 | 22.2%8 / 36 requests | failing | 09 Nov 2026, 02:13 UTC |
| Sable via NanoGPT |
DeepSeek V4.1 Flash 1017 | heretic | NATIVE | 512,000 | $0.303 | $0.173 | $0.693 | 110.7 | 3.8 | 8.9 | 97.2%35 / 36 requests | ok | 11 Nov 2026, 10:15 UTC |
| OpenRoutermorph/fp8 |
DeepSeek V4.1 Flash 1017 | none | FP8 | 1,048,576 | $0.117 | $0.022 | $0.400 | 96.8 | 0.7 | 6.5 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:13 UTC |
| Venice |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.578 | $0.330 | $1.320 | 180.6 | 0.4 | 3.5 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:07 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- Experiential Labs · dealignai
- Heabsy via NanoGPT
- OrcaRouter via NanoGPT
- Sable via NanoGPT
- OpenRouter · morph/fp8
- Venice
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| Experiential Labsdealignai |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.394 | $0.225 | $0.900 | 223.1 | 0.5 | 2.8 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:08 UTC |
| Heabsy via NanoGPT |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.610 | $0.350 | $1.390 | — | — | — | 0.0%0 / 36 requests | failing | 08 Nov 2026, 10:13 UTC |
| OrcaRouter via NanoGPT |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $1.050 | $0.840 | $1.680 | 152.4 | 1.5 | 4.7 | 22.2%8 / 36 requests | failing | 09 Nov 2026, 02:13 UTC |
| Sable via NanoGPT |
DeepSeek V4.1 Flash 1017 | heretic | NATIVE | 512,000 | $0.303 | $0.173 | $0.693 | 117.2 | 3.7 | 8.2 | 94.4%34 / 36 requests | ok | 11 Nov 2026, 10:15 UTC |
| OpenRoutermorph/fp8 |
DeepSeek V4.1 Flash 1017 | none | FP8 | 1,048,576 | $0.117 | $0.022 | $0.400 | 87.5 | 0.8 | 7.2 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:13 UTC |
| Venice |
DeepSeek V4.1 Flash 1017 | uncensored | Unknown | 1,000,000 | $0.578 | $0.330 | $1.320 | 183.8 | 0.6 | 3.4 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:07 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- Experiential Labs · dealignai
- Heabsy via NanoGPT
- OrcaRouter via NanoGPT
- Sable via NanoGPT
- OpenRouter · morph/fp8
- Venice
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
 abliteration.aiAbliterated Large Flash |
DeepSeek V4.1 Flash (undated derivative) | abliterated | Unknown | 1,048,576 | $1.250 | $1.000 | $2.000 | 191.2 | 1.1 | 4.0 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:07 UTC |
| abliteration.ai via NanoGPT |
DeepSeek V4.1 Flash (undated derivative) | abliterated | FP8 | 1,000,000 | $1.312 | $1.050 | $2.100 | 191.0 | 2.1 | 4.5 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:15 UTC |
| Venice via NanoGPT |
DeepSeek V4.1 Flash (undated derivative) | uncensored | FP8 | 524,288 | $0.608 | $0.347 | $1.390 | — | — | — | 0.0%0 / 72 requests | failing | 08 Nov 2026, 02:13 UTC |
 Refuseless |
DeepSeek V4.1 Flash (undated derivative) | abliterated | FP8 | 1,048,576 | $0.203 | $0.120 | $0.450 | 141.9 | 2.0 | 6.0 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:10 UTC |
 Yeta Labs |
DeepSeek V4.1 Flash (undated derivative) | uncensored | FP8 | 1,000,000 | $0.147 | $0.080 | $0.350 | 128.7 | 1.7 | 5.7 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:05 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- abliteration.ai · Abliterated Large Flash
- abliteration.ai via NanoGPT
- Venice via NanoGPT
- Refuseless
- Yeta Labs
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| abliteration.aiAbliterated Large Flash |
DeepSeek V4.1 Flash (undated derivative) | abliterated | Unknown | 1,048,576 | $1.250 | $1.000 | $2.000 | 209.6 | 1.1 | 3.8 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:07 UTC |
| abliteration.ai via NanoGPT |
DeepSeek V4.1 Flash (undated derivative) | abliterated | FP8 | 1,000,000 | $1.312 | $1.050 | $2.100 | 179.0 | 2.0 | 5.1 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:15 UTC |
| Venice via NanoGPT |
DeepSeek V4.1 Flash (undated derivative) | uncensored | FP8 | 524,288 | $0.608 | $0.347 | $1.390 | — | — | — | 0.0%0 / 36 requests | failing | 08 Nov 2026, 02:13 UTC |
| Refuseless |
DeepSeek V4.1 Flash (undated derivative) | abliterated | FP8 | 1,048,576 | $0.203 | $0.120 | $0.450 | 143.9 | 1.9 | 5.9 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:10 UTC |
| Yeta Labs |
DeepSeek V4.1 Flash (undated derivative) | uncensored | FP8 | 1,000,000 | $0.147 | $0.080 | $0.350 | 130.8 | 1.6 | 5.9 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:05 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- abliteration.ai · Abliterated Large Flash
- abliteration.ai via NanoGPT
- Venice via NanoGPT
- Refuseless
- Yeta Labs
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| abliteration.aiAbliterated Large Flash |
DeepSeek V4.1 Flash (undated derivative) | abliterated | Unknown | 1,048,576 | $1.250 | $1.000 | $2.000 | 174.2 | 1.2 | 4.1 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:07 UTC |
| abliteration.ai via NanoGPT |
DeepSeek V4.1 Flash (undated derivative) | abliterated | FP8 | 1,000,000 | $1.312 | $1.050 | $2.100 | 216.9 | 2.1 | 4.4 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:15 UTC |
| Venice via NanoGPT |
DeepSeek V4.1 Flash (undated derivative) | uncensored | FP8 | 524,288 | $0.608 | $0.347 | $1.390 | — | — | — | 0.0%0 / 36 requests | failing | 08 Nov 2026, 02:13 UTC |
| Refuseless |
DeepSeek V4.1 Flash (undated derivative) | abliterated | FP8 | 1,048,576 | $0.203 | $0.120 | $0.450 | 137.7 | 2.0 | 6.4 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:10 UTC |
| Yeta Labs |
DeepSeek V4.1 Flash (undated derivative) | uncensored | FP8 | 1,000,000 | $0.147 | $0.080 | $0.350 | 128.7 | 1.8 | 5.5 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:05 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- abliteration.ai · Abliterated Large Flash
- abliteration.ai via NanoGPT
- Venice via NanoGPT
- Refuseless
- Yeta Labs
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| Experiential Labsdealignai |
GLM 5.3 Flash | uncensored | Unknown | 1,000,000 | $0.237 | $0.150 | $0.500 | 82.4 | 1.0 | 7.7 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:09 UTC |
 Arnict via NanoGPT |
GLM 5.3 Flash | uncensored | NVFP4 | 1,000,000 | $0.230 | $0.131 | $0.525 | 165.4 | 1.4 | 4.5 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:14 UTC |
| Heabsy via NanoGPT |
GLM 5.3 Flash | uncensored | NVFP4 | 1,000,000 | $0.350 | $0.200 | $0.800 | 181.1 | 1.1 | 3.9 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:14 UTC |
| OrcaRouter via NanoGPT |
GLM 5.3 Flash | uncensored | Unknown | 1,000,000 | $1.050 | $0.840 | $1.680 | 48.8 | 2.1 | 12.6 | 22.2%16 / 72 requests | failing | 09 Nov 2026, 02:12 UTC |
| OpenRouterz-ai/fp8 |
GLM 5.3 Flash | none | FP8 | 1,048,576 | $0.237 | $0.150 | $0.500 | 47.3 | 1.3 | 13.0 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:12 UTC |
| Refuseless |
GLM 5.3 Flash | abliterated | NVFP4 | 1,048,576 | $0.215 | $0.120 | $0.500 | 47.3 | 1.6 | 13.0 | 98.6%71 / 72 requests | ok | 11 Nov 2026, 10:12 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- Experiential Labs · dealignai
- Arnict via NanoGPT
- Heabsy via NanoGPT
- OrcaRouter via NanoGPT
- OpenRouter · z-ai/fp8
- Refuseless
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| Experiential Labsdealignai |
GLM 5.3 Flash | uncensored | Unknown | 1,000,000 | $0.237 | $0.150 | $0.500 | 98.0 | 1.1 | 7.7 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:09 UTC |
| Arnict via NanoGPT |
GLM 5.3 Flash | uncensored | NVFP4 | 1,000,000 | $0.230 | $0.131 | $0.525 | 165.9 | 1.4 | 4.4 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:14 UTC |
| Heabsy via NanoGPT |
GLM 5.3 Flash | uncensored | NVFP4 | 1,000,000 | $0.350 | $0.200 | $0.800 | 194.1 | 1.1 | 3.9 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:14 UTC |
| OrcaRouter via NanoGPT |
GLM 5.3 Flash | uncensored | Unknown | 1,000,000 | $1.050 | $0.840 | $1.680 | 48.8 | 2.1 | 12.9 | 22.2%8 / 36 requests | failing | 09 Nov 2026, 02:12 UTC |
| OpenRouterz-ai/fp8 |
GLM 5.3 Flash | none | FP8 | 1,048,576 | $0.237 | $0.150 | $0.500 | 48.7 | 1.4 | 12.3 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:12 UTC |
| Refuseless |
GLM 5.3 Flash | abliterated | NVFP4 | 1,048,576 | $0.215 | $0.120 | $0.500 | 46.7 | 1.7 | 13.3 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:12 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- Experiential Labs · dealignai
- Arnict via NanoGPT
- Heabsy via NanoGPT
- OrcaRouter via NanoGPT
- OpenRouter · z-ai/fp8
- Refuseless
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| Experiential Labsdealignai |
GLM 5.3 Flash | uncensored | Unknown | 1,000,000 | $0.237 | $0.150 | $0.500 | 74.0 | 0.9 | 7.9 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:09 UTC |
| Arnict via NanoGPT |
GLM 5.3 Flash | uncensored | NVFP4 | 1,000,000 | $0.230 | $0.131 | $0.525 | 161.4 | 1.5 | 4.7 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:14 UTC |
| Heabsy via NanoGPT |
GLM 5.3 Flash | uncensored | NVFP4 | 1,000,000 | $0.350 | $0.200 | $0.800 | 178.1 | 1.1 | 3.8 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:14 UTC |
| OrcaRouter via NanoGPT |
GLM 5.3 Flash | uncensored | Unknown | 1,000,000 | $1.050 | $0.840 | $1.680 | 50.8 | 2.1 | 12.6 | 22.2%8 / 36 requests | failing | 09 Nov 2026, 02:12 UTC |
| OpenRouterz-ai/fp8 |
GLM 5.3 Flash | none | FP8 | 1,048,576 | $0.237 | $0.150 | $0.500 | 46.9 | 1.3 | 13.1 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:12 UTC |
| Refuseless |
GLM 5.3 Flash | abliterated | NVFP4 | 1,048,576 | $0.215 | $0.120 | $0.500 | 54.0 | 1.6 | 10.3 | 97.2%35 / 36 requests | ok | 11 Nov 2026, 10:12 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- Experiential Labs · dealignai
- Arnict via NanoGPT
- Heabsy via NanoGPT
- OrcaRouter via NanoGPT
- OpenRouter · z-ai/fp8
- Refuseless
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| abliteration.aiAbliterated Large v2 |
GLM 5.3 | abliterated | Unknown | 1,000,000 | $3.500 | $3.000 | $5.000 | 182.5 | 1.2 | 4.0 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:10 UTC |
| abliteration.ai via NanoGPT |
GLM 5.3 | abliterated | FP8 | 1,000,000 | $3.675 | $3.150 | $5.250 | 162.4 | 2.2 | 5.3 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:15 UTC |
| Arnict via NanoGPT |
GLM 5.3 | uncensored | NVFP4 | 1,000,000 | $1.302 | $1.040 | $2.090 | 53.0 | 2.7 | 12.3 | 95.8%69 / 72 requests | ok | 11 Nov 2026, 10:15 UTC |
| OpenRouterz-ai/fp8 |
GLM 5.3 | none | FP8 | 1,048,576 | $2.150 | $1.400 | $4.400 | 67.5 | 1.6 | 9.3 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:09 UTC |
| Refuseless |
GLM 5.3 | abliterated | NVFP4 | 1,048,576 | $1.250 | $1.000 | $2.000 | 74.1 | 1.8 | 9.2 | 97.2%70 / 72 requests | ok | 11 Nov 2026, 10:12 UTC |
| VeniceAbliterated Large v2 via abliteration.ai |
GLM 5.3 | abliterated | Unknown | 1,000,000 | $3.500 | $3.000 | $5.000 | 229.8 | 1.5 | 3.8 | 100.0%72 / 72 requests | ok | 11 Nov 2026, 10:10 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- abliteration.ai · Abliterated Large v2
- abliteration.ai via NanoGPT
- Arnict via NanoGPT
- OpenRouter · z-ai/fp8
- Refuseless
- Venice · Abliterated Large v2 via abliteration.ai
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| abliteration.aiAbliterated Large v2 |
GLM 5.3 | abliterated | Unknown | 1,000,000 | $3.500 | $3.000 | $5.000 | 182.5 | 1.2 | 4.1 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:10 UTC |
| abliteration.ai via NanoGPT |
GLM 5.3 | abliterated | FP8 | 1,000,000 | $3.675 | $3.150 | $5.250 | 174.8 | 2.1 | 5.1 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:15 UTC |
| Arnict via NanoGPT |
GLM 5.3 | uncensored | NVFP4 | 1,000,000 | $1.302 | $1.040 | $2.090 | 55.5 | 3.0 | 12.0 | 94.4%34 / 36 requests | ok | 11 Nov 2026, 10:15 UTC |
| OpenRouterz-ai/fp8 |
GLM 5.3 | none | FP8 | 1,048,576 | $2.150 | $1.400 | $4.400 | 67.5 | 1.6 | 9.1 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:09 UTC |
| Refuseless |
GLM 5.3 | abliterated | NVFP4 | 1,048,576 | $1.250 | $1.000 | $2.000 | 76.5 | 1.5 | 8.7 | 97.2%35 / 36 requests | ok | 11 Nov 2026, 10:12 UTC |
| VeniceAbliterated Large v2 via abliteration.ai |
GLM 5.3 | abliterated | Unknown | 1,000,000 | $3.500 | $3.000 | $5.000 | 224.3 | 1.4 | 3.8 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:10 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- abliteration.ai · Abliterated Large v2
- abliteration.ai via NanoGPT
- Arnict via NanoGPT
- OpenRouter · z-ai/fp8
- Refuseless
- Venice · Abliterated Large v2 via abliteration.ai
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗
Output speed
tok/s · Higher is better · p50
Reported completion tokens after the first chunk, divided by streaming time. Samples without reported usage are excluded. Methodology ↗
Time to first token
s · Lower is better · p50
Request sent to first non-empty content, reasoning or tool-argument delta. Includes network time from the selected region. Methodology ↗
Time to 500 tokens
s · Lower is better · p50
Per request: TTFT + 500 ÷ output tok/s. Percentiles are computed after this calculation; this is a derived response time. Methodology ↗
Blended price
$ / 1M tokens · Lower is better
Declared price: (3 × input + output) ÷ 4. Excludes cache discounts. Prices are not measured billing receipts. Methodology ↗
| abliteration.aiAbliterated Large v2 |
GLM 5.3 | abliterated | Unknown | 1,000,000 | $3.500 | $3.000 | $5.000 | 193.1 | 1.2 | 3.9 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:10 UTC |
| abliteration.ai via NanoGPT |
GLM 5.3 | abliterated | FP8 | 1,000,000 | $3.675 | $3.150 | $5.250 | 145.1 | 2.4 | 5.3 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:15 UTC |
| Arnict via NanoGPT |
GLM 5.3 | uncensored | NVFP4 | 1,000,000 | $1.302 | $1.040 | $2.090 | 52.6 | 2.7 | 12.7 | 97.2%35 / 36 requests | ok | 11 Nov 2026, 10:15 UTC |
| OpenRouterz-ai/fp8 |
GLM 5.3 | none | FP8 | 1,048,576 | $2.150 | $1.400 | $4.400 | 67.3 | 1.5 | 9.4 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:09 UTC |
| Refuseless |
GLM 5.3 | abliterated | NVFP4 | 1,048,576 | $1.250 | $1.000 | $2.000 | 68.8 | 2.0 | 9.3 | 97.2%35 / 36 requests | ok | 11 Nov 2026, 10:12 UTC |
| VeniceAbliterated Large v2 via abliteration.ai |
GLM 5.3 | abliterated | Unknown | 1,000,000 | $3.500 | $3.000 | $5.000 | 235.0 | 1.5 | 3.8 | 100.0%36 / 36 requests | ok | 11 Nov 2026, 10:10 UTC |
p50 = median. Missing speed or latency values mean fewer than 8 usable successes. Success rate includes failed requests. Status and last success describe the route across all regions. Precision and context are provider declarations; different modifications are different products. Measurement details →
Output speed vs. price
Median output speed and declared blended price
- abliteration.ai · Abliterated Large v2
- abliteration.ai via NanoGPT
- Arnict via NanoGPT
- OpenRouter · z-ai/fp8
- Refuseless
- Venice · Abliterated Large v2 via abliteration.ai
Each point is one route. Shading divides the displayed axes at their midpoints; it is not a quality rating. Blended price uses a 3:1 input/output mix. Routes without usable speed samples are omitted. Methodology ↗