1gpt-5.4 xhighLeader
GPT-5.4 · xhigh
Raw avg
89.0
P10 floor
79.9
Cost / call
$0.07
Calls
n=50
salesevals.com/Evaluated Jul 1, 2026
Which models know sales?
26 model configurations coach GPT- and Sonnet-generated synthetic sales calls with hidden ground truth. A judge scores each coaching note from 0–100 on whether it found the real strengths, flaws, and next moves.
- Calls
- 50
- Models
- 26
- Evaluations
- 1300
- Benchmark
- 86.2
50 calls · 1300 evaluationsRank: Sales coaching benchmarkAll available runsBuild-time static dataEvals completed Jul 1, 2026
Ranked by benchmark score
Leaderboard
Weighted for needle recall, sales instinct, prioritization, technical accuracy, false-positive control, and raw score across 1300 evaluations.
89.3
2gpt-5.4 high
GPT-5.4 · high
Raw avg
89.0
P10 floor
81.9
Cost / call
$0.07
Calls
n=50
3gpt-5.5 medium
GPT-5.5 · medium
Raw avg
88.8
P10 floor
82.0
Cost / call
$0.17
Calls
n=50
4gpt-5.5 xhigh
GPT-5.5 · xhigh
Raw avg
89.0
P10 floor
78.5
Cost / call
$0.18
Calls
n=50
5gpt-5.5 high
GPT-5.5 · high
Raw avg
88.6
P10 floor
79.4
Cost / call
$0.18
Calls
n=50
6gpt-5.4 medium
GPT-5.4 · medium
Raw avg
88.3
P10 floor
80.1
Cost / call
$0.07
Calls
n=50
7gpt-5.5 none
GPT-5.5 · none
Raw avg
88.1
P10 floor
76.6
Cost / call
$0.17
Calls
n=50
8gpt-5.5 low
GPT-5.5 · low
Raw avg
87.7
P10 floor
78.3
Cost / call
$0.17
Calls
n=50
9fable 5 high
Claude Fable 5 · high
Raw avg
87.5
P10 floor
77.0
Cost / call
$0.32
Calls
n=50
10gpt-5.4 low
GPT-5.4 · low
Raw avg
87.4
P10 floor
78.3
Cost / call
$0.07
Calls
n=50
11gpt-5.4 none
GPT-5.4 · none
Raw avg
87.4
P10 floor
78.2
Cost / call
$0.07
Calls
n=50
12opus 4.7 max
Claude Opus 4.7 · max
Raw avg
87.3
P10 floor
75.0
Cost / call
$0.16
Calls
n=50
13opus 4.7 high
Claude Opus 4.7 · high
Raw avg
86.8
P10 floor
77.8
Cost / call
$0.13
Calls
n=50
14opus 4.8 medium
Claude Opus 4.8 · medium
Raw avg
85.8
P10 floor
72.1
Cost / call
$0.11
Calls
n=50
15opus 4.7 medium
Claude Opus 4.7 · medium
Raw avg
85.6
P10 floor
74.0
Cost / call
$0.11
Calls
n=50
16opus 4.7 xhigh
Claude Opus 4.7 · xhigh
Raw avg
85.6
P10 floor
73.1
Cost / call
$0.14
Calls
n=50
17opus 4.7 low
Claude Opus 4.7 · low
Raw avg
85.6
P10 floor
75.6
Cost / call
$0.10
Calls
n=50
18opus 4.8 max
Claude Opus 4.8 · max
Raw avg
85.4
P10 floor
72.9
Cost / call
$0.15
Calls
n=50
19opus 4.8 xhigh
Claude Opus 4.8 · xhigh
Raw avg
85.2
P10 floor
73.0
Cost / call
$0.13
Calls
n=50
20opus 4.8 high
Claude Opus 4.8 · high
Raw avg
84.9
P10 floor
68.8
Cost / call
$0.11
Calls
n=50
21sonnet 4.6
Claude Sonnet 4.6 · default
Raw avg
84.6
P10 floor
71.9
Cost / call
$0.10
Calls
n=50
22sonnet 5
Claude Sonnet 5 · default
Raw avg
84.6
P10 floor
71.6
Cost / call
$0.05
Calls
n=50
23opus 4.8 low
Claude Opus 4.8 · low
Raw avg
84.0
P10 floor
67.2
Cost / call
$0.10
Calls
n=50
24glm 5.2
GLM 5.2 · default
Raw avg
84.0
P10 floor
71.3
Cost / call
$0.03
Calls
n=50
25deepseek v4 pro
DeepSeek V4 Pro · default
Raw avg
83.5
P10 floor
69.1
Cost / call
$0.0047
Calls
n=50
26gemini 3.1 pro previewTrailing
Gemini 3.1 Pro Preview · default
Raw avg
78.9
P10 floor
66.6
Cost / call
$0.03
Calls
n=50
Filtered benchmark86.2
| # | Model | Benchmark | Raw avg | P10 floor | Cost / call | Distribution | n |
|---|---|---|---|---|---|---|---|
| 1 | gpt-5.4 xhighLeader GPT-5.4 · xhigh | 89.3 | 89.0 | 79.9 | $0.07 | 50 | |
| 2 | gpt-5.4 high GPT-5.4 · high | 89.2 | 89.0 | 81.9 | $0.07 | 50 | |
| 3 | gpt-5.5 medium GPT-5.5 · medium | 89.1 | 88.8 | 82.0 | $0.17 | 50 | |
| 4 | gpt-5.5 xhigh GPT-5.5 · xhigh | 89.1 | 89.0 | 78.5 | $0.18 | 50 | |
| 5 | gpt-5.5 high GPT-5.5 · high | 88.9 | 88.6 | 79.4 | $0.18 | 50 | |
| 6 | gpt-5.4 medium GPT-5.4 · medium | 88.4 | 88.3 | 80.1 | $0.07 | 50 | |
| 7 | gpt-5.5 none GPT-5.5 · none | 88.3 | 88.1 | 76.6 | $0.17 | 50 | |
| 8 | gpt-5.5 low GPT-5.5 · low | 87.8 | 87.7 | 78.3 | $0.17 | 50 | |
| 9 | fable 5 high Claude Fable 5 · high | 87.7 | 87.5 | 77.0 | $0.32 | 50 | |
| 10 | gpt-5.4 low GPT-5.4 · low | 87.5 | 87.4 | 78.3 | $0.07 | 50 | |
| 11 | gpt-5.4 none GPT-5.4 · none | 87.5 | 87.4 | 78.2 | $0.07 | 50 | |
| 12 | opus 4.7 max Claude Opus 4.7 · max | 87.2 | 87.3 | 75.0 | $0.16 | 50 | |
| 13 | opus 4.7 high Claude Opus 4.7 · high | 86.6 | 86.8 | 77.8 | $0.13 | 50 | |
| 14 | opus 4.8 medium Claude Opus 4.8 · medium | 85.6 | 85.8 | 72.1 | $0.11 | 50 | |
| 15 | opus 4.7 medium Claude Opus 4.7 · medium | 85.5 | 85.6 | 74.0 | $0.11 | 50 | |
| 16 | opus 4.7 xhigh Claude Opus 4.7 · xhigh | 85.5 | 85.6 | 73.1 | $0.14 | 50 | |
| 17 | opus 4.7 low Claude Opus 4.7 · low | 85.5 | 85.6 | 75.6 | $0.10 | 50 | |
| 18 | opus 4.8 max Claude Opus 4.8 · max | 85.4 | 85.4 | 72.9 | $0.15 | 50 | |
| 19 | opus 4.8 xhigh Claude Opus 4.8 · xhigh | 85.3 | 85.2 | 73.0 | $0.13 | 50 | |
| 20 | opus 4.8 high Claude Opus 4.8 · high | 84.9 | 84.9 | 68.8 | $0.11 | 50 | |
| 21 | sonnet 4.6 Claude Sonnet 4.6 · default | 84.5 | 84.6 | 71.9 | $0.10 | 50 | |
| 22 | sonnet 5 Claude Sonnet 5 · default | 84.3 | 84.6 | 71.6 | $0.05 | 50 | |
| 23 | opus 4.8 low Claude Opus 4.8 · low | 83.7 | 84.0 | 67.2 | $0.10 | 50 | |
| 24 | glm 5.2 GLM 5.2 · default | 83.6 | 84.0 | 71.3 | $0.03 | 50 | |
| 25 | deepseek v4 pro DeepSeek V4 Pro · default | 83.1 | 83.5 | 69.1 | $0.0047 | 50 | |
| 26 | gemini 3.1 pro previewTrailing Gemini 3.1 Pro Preview · default | 78.7 | 78.9 | 66.6 | $0.03 | 50 | |
| Filtered benchmark | 86.2 | ||||||
Top model
89.3benchmark
gpt-5.4 xhigh
89.0 raw avg · 79.9 p10 floor · $0.07/call
Hardest call
65.5avg
AnthropicExxonMobil
ExxonMobil AI governance and safety review for energy operations with Anthropic
Product demomixedSonnet-generated
Reasoning effort
+1.8pts
GPT-5.4: none → xhigh
87.5 → 89.3 as effort scales
How this benchmark was made
Methodology
The calls are synthetic, immutable benchmark cases with hidden coaching ground truth.
01
Generate cases
Each case starts with two companies, a call type, duration, quality target, web research, and role notes.
02
Write turn by turn
Half the current calls come from the original GPT-based generator and half from a Claude Sonnet 4.6 generator. Both are written one speaker turn at a time.
03
Judge semantically
Coach models see only the visible case. A judge model sees hidden ground truth, scores eight dimensions, and rolls the sales-critical axes into the benchmark ranking.
- Calls
- 50
- Model configs
- 26
- Judged runs
- 1300
- Origins
- 2
- Axes
- 8
50 sales calls in the benchmark
Browse the calls
Pick a call to see the hidden answer key and model scores.
Discovery95.6
CollibraBerkshire Hathaway
Berkshire Hathaway Data governance discovery across decentralized business units with Collibra
flawed·GPT-generated·33mEasiest
Competitive displacement94.4
StripePave
Pave Pricing and packaging objection call with Stripe
flawed·GPT-generated·18m
Discovery94.0
AtlassianDelta Air Lines
Delta Air Lines Enterprise discovery for service management modernization with Atlassian
flawed·GPT-generated·31m
Discovery93.9
VercelMercury
Mercury First discovery for frontend platform consolidation with Vercel
flawed·GPT-generated·22m
Discovery93.9
WorkdayMcKesson
McKesson HR transformation qualification and stakeholder mapping with Workday
flawed·Sonnet-generated·27m
Renewal save93.8
TwilioThe Home Depot
The Home Depot Renewal save call after usage and support concerns with Twilio
flawed·GPT-generated·42m
Product demo93.3
MongoDBWayfair
Wayfair Integration deep dive for catalog modernization with MongoDB
excellent·GPT-generated·58m
Product demo92.9
Palo Alto NetworksApple
Apple Technical security review for zero trust architecture with Palo Alto Networks
excellent·GPT-generated·66m
QBR92.5
AmplitudeDuolingo
Duolingo Renewal QBR and expansion planning with Amplitude
excellent·GPT-generated·52m
Discovery91.7
WorkdayMcKesson
McKesson HR transformation qualification and stakeholder mapping with Workday
flawed·GPT-generated·27m
Discovery91.7
OpenAICVS Health
CVS Health AI contact-center transformation discovery with OpenAI
excellent·GPT-generated·61m
Discovery91.7
GitHubRippling
Rippling Product-led expansion discovery for developer workflow with GitHub
excellent·GPT-generated·41m
Competitive displacement91.3
CloudflareCanva
Canva Competitive displacement discovery for edge security with Cloudflare
flawed·Sonnet-generated·47m
Discovery90.8
VercelMercury
Mercury First discovery for frontend platform consolidation with Vercel
flawed·Sonnet-generated·22m
Product demo90.2
CrowdStrikeTarget
Target Security architecture review for endpoint consolidation with CrowdStrike
excellent·GPT-generated·63m
Competitive displacement90.2
StripePave
Pave Pricing and packaging objection call with Stripe
flawed·Sonnet-generated·18m
Product demo90.0
DatadogLinear
Linear Technical demo for observability and incident response with Datadog
excellent·GPT-generated·34m
Product demo89.9
AnthropicExxonMobil
ExxonMobil AI governance and safety review for energy operations with Anthropic
mixed·GPT-generated·39m
Product demo89.7
ElasticJPMorgan Chase
JPMorgan Chase Technical workshop for search and observability consolidation with Elastic
excellent·GPT-generated·74m
Product demo89.3
MongoDBWayfair
Wayfair Integration deep dive for catalog modernization with MongoDB
excellent·Sonnet-generated·58m
Discovery89.3
HashiCorpAmazon
Amazon Cloud operating model discussion for internal platform teams with HashiCorp
flawed·GPT-generated·26m
Product demo88.9
MicrosoftCostco Wholesale
Costco Wholesale Proof-of-concept readout for analytics and productivity workflow with Microsoft
mixed·Sonnet-generated·55m
Discovery88.6
NVIDIAWalmart
Walmart Executive discovery for AI infrastructure and store operations with NVIDIA
excellent·GPT-generated·57m
Competitive displacement88.2
ServiceNowFord Motor Company
Ford Motor Company Procurement negotiation for workflow automation with ServiceNow
mixed·GPT-generated·35m
Product demo88.0
CrowdStrikeTarget
Target Security architecture review for endpoint consolidation with CrowdStrike
excellent·Sonnet-generated·63m
Discovery88.0
GitHubRippling
Rippling Product-led expansion discovery for developer workflow with GitHub
excellent·Sonnet-generated·41m
Discovery88.0
OpenAICVS Health
CVS Health AI contact-center transformation discovery with OpenAI
excellent·Sonnet-generated·61m
Product demo86.7
SnowflakeToast
Toast Data platform proof-of-concept kickoff with Snowflake
flawed·GPT-generated·44m
Discovery85.8
NVIDIAWalmart
Walmart Executive discovery for AI infrastructure and store operations with NVIDIA
excellent·Sonnet-generated·57m
Competitive displacement85.2
CloudflareCanva
Canva Competitive displacement discovery for edge security with Cloudflare
flawed·GPT-generated·47m
Discovery84.8
AtlassianDelta Air Lines
Delta Air Lines Enterprise discovery for service management modernization with Atlassian
flawed·Sonnet-generated·31m
Discovery84.8
HashiCorpAmazon
Amazon Cloud operating model discussion for internal platform teams with HashiCorp
flawed·Sonnet-generated·26m
QBR84.7
OktaSweetgreen
Sweetgreen Executive alignment for identity modernization with Okta
mixed·Sonnet-generated·38m
QBR84.3
OktaSweetgreen
Sweetgreen Executive alignment for identity modernization with Okta
mixed·GPT-generated·38m
Product demo84.1
FigmaThe Walt Disney Company
The Walt Disney Company Design collaboration demo with brand and asset workflow discussion with Figma
mixed·GPT-generated·49m
Renewal save83.9
SalesforceUnitedHealth Group
UnitedHealth Group Healthcare CRM expansion objection handling with Salesforce
mixed·GPT-generated·46m
Product demo83.5
SnykRunway
Runway Security review before developer-tool rollout with Snyk
mixed·Sonnet-generated·29m
Product demo83.0
SnykRunway
Runway Security review before developer-tool rollout with Snyk
mixed·GPT-generated·29m
Renewal save81.8
TwilioThe Home Depot
The Home Depot Renewal save call after usage and support concerns with Twilio
flawed·Sonnet-generated·42m
Renewal save81.5
SalesforceUnitedHealth Group
UnitedHealth Group Healthcare CRM expansion objection handling with Salesforce
mixed·Sonnet-generated·46m
Product demo81.0
DatadogLinear
Linear Technical demo for observability and incident response with Datadog
excellent·Sonnet-generated·34m
QBR80.5
AmplitudeDuolingo
Duolingo Renewal QBR and expansion planning with Amplitude
excellent·Sonnet-generated·52m
Product demo80.1
FigmaThe Walt Disney Company
The Walt Disney Company Design collaboration demo with brand and asset workflow discussion with Figma
mixed·Sonnet-generated·49m
Product demo79.1
Palo Alto NetworksApple
Apple Technical security review for zero trust architecture with Palo Alto Networks
excellent·Sonnet-generated·66m
Competitive displacement77.3
ServiceNowFord Motor Company
Ford Motor Company Procurement negotiation for workflow automation with ServiceNow
mixed·Sonnet-generated·35m
Product demo76.7
MicrosoftCostco Wholesale
Costco Wholesale Proof-of-concept readout for analytics and productivity workflow with Microsoft
mixed·GPT-generated·55m
Product demo76.5
SnowflakeToast
Toast Data platform proof-of-concept kickoff with Snowflake
flawed·Sonnet-generated·44m
Product demo71.3
ElasticJPMorgan Chase
JPMorgan Chase Technical workshop for search and observability consolidation with Elastic
excellent·Sonnet-generated·74m
Discovery70.3
CollibraBerkshire Hathaway
Berkshire Hathaway Data governance discovery across decentralized business units with Collibra
flawed·Sonnet-generated·33m
Product demo65.5
AnthropicExxonMobil
ExxonMobil AI governance and safety review for energy operations with Anthropic
mixed·Sonnet-generated·39mHardest
By model family
Does spending more on reasoning help?
For families with multiple reasoning settings, this shows the benchmark score at each level.
Claude Opus 4.7
+1.7 pts
low
85.5
baseline
medium
85.5
+0.0 pts
high
86.6
+1.1 pts
xhigh
85.5
+0.0 pts
max
87.2
+1.7 pts
Claude Opus 4.8
+1.7 pts
low
83.7
baseline
medium
85.6
+1.8 pts
high
84.9
+1.1 pts
xhigh
85.3
+1.6 pts
max
85.4
+1.7 pts
GPT-5.4
+1.8 pts
none
87.5
baseline
low
87.5
+0.0 pts
medium
88.4
+1.0 pts
high
89.2
+1.7 pts
xhigh
89.3
+1.8 pts
GPT-5.5
+0.8 pts
none
88.3
baseline
low
87.8
-0.5 pts
medium
89.1
+0.8 pts
high
88.9
+0.6 pts
xhigh
89.1
+0.8 pts